CN118012432B - Terminal intelligent experience realization method, device, medium and equipment - Google Patents

Abstract

The application provides an end intelligent experience realization method, device, medium and equipment, which are used for sensing, deciding and disposing user experience at the front end aiming at front end codes suitable for multiple types of clients, wherein the method comprises the following steps: starting a client, sending an intelligent experience configuration request to a server, acquiring intelligent experience configuration of the server and storing the intelligent experience configuration to a local; running an end-intelligence suite in front-end code, comprising: starting a perception engine in the intelligent kit of the terminal to acquire target data in real time; triggering a decision engine in the terminal intelligent suite, and carrying out experience decision aiming at target data according to the terminal intelligent experience configuration; and triggering a corresponding disposal scheme according to the experience decision result by a contact engine in the intelligent kit of the running end. The embodiment of the application can realize front-end intellectualization.

Description

Method, device, medium and equipment for realizing terminal intelligent experience

Technical Field

The present application relates to the field of application development technology, and in particular to a method, device, medium and equipment for implementing a terminal intelligent experience.

Background technique

It is an important part of modern life for users to work, communicate, shop, etc. through various clients (such as APP, web pages, websites, mini-programs, etc.) on terminals (such as smartphones). The instant experience of users when using the client may involve a series of functions such as interface design, page loading, function response, usability, fluency, user feedback, etc. How to develop and improve applications based on improving user experience is a technical issue that technicians in this field need to consider.

Summary of the invention

In view of this, the present application provides a method, device, medium and electronic device for realizing end-to-end intelligent experience, the main purpose of which is to improve user experience by realizing end-to-end intelligence.

According to one aspect of the present application, a method for implementing a terminal intelligent experience is provided, which is used to perceive, decide and handle user experience at the front end for front-end codes applicable to multiple types of clients, and the method is applied to one client of any type of client among the multiple types of clients, including:

Start the client, send a request for the client intelligent experience configuration to the server, obtain the client intelligent experience configuration and store it locally;

Running the end intelligence suite in the front-end code includes: starting the perception engine in the end intelligence suite to obtain target data in real time; triggering the decision engine in the end intelligence suite to make experience decisions for the target data based on the end intelligence experience configuration; and running the contact engine in the end intelligence suite to trigger corresponding disposal plans based on the experience decision results.

In one implementation, the treatment scheme includes: an experience monitoring treatment scheme, an experience layering treatment scheme, and/or an experience enhancement treatment scheme;

The experience monitoring and handling scheme includes: when a complex event is determined to occur according to the experience decision, real-time message push and/or page jump operation is performed in the interface in a UI interactive manner;

The experience layering solution includes: when the device operation state is determined to be low or poor according to the experience decision, performing interface loading optimization and/or page content dynamic adjustment;

The experience enhancement solution includes: obtaining a user's future behavior trajectory based on the experience decision, determining sub-package resources or page resources that the user may access, and preloading the sub-package resources or page resources.

In one implementation, the method further includes:

Set different handling plans to correspond to different priorities, and determine the execution logic relationship between different handling plans according to the priorities; among them, set the experience stratification handling plan to correspond to the first priority; according to the processing results of the experience stratification handling plan, determine whether to execute the experience monitoring handling plan and/or the experience enhancement handling plan.

In one implementation, the method further includes:

Use white box models or black box models to make experience decisions for experience monitoring, experience layering, or experience enhancement, and make preliminary decisions based on the white box model, and determine whether to use the black box model to make further refined decisions based on the preliminary decision results of the white box model.

In one implementation, the method further includes:

Setting the experience layering model corresponding to the experience layering decision to a white box model mode, running the experience layering model in the white box model mode during the cold start phase of the client, and determining whether to continue to execute the terminal intelligence suite or shut down the terminal intelligence suite according to the prediction result of the experience layering model;

The full-link traffic flow prediction model corresponding to the experience enhancement decision is set to a black box model mode, and during the client hot start phase, the full-link traffic flow prediction model of the black box model mode is run.

In one implementation, the terminal intelligent experience configuration includes complex event configuration information, and the complex event configuration information includes event configuration information and timing configuration information of at least one target complex event;

The perception engine in the startup end intelligent suite acquires target data in real time, including: starting the perception engine, capturing embedded point data based on the unified event stream processing tool in the front-end code, and determining whether the target complex event corresponding to the event configuration information is triggered according to the embedded point data;

The decision engine in the trigger end intelligent suite makes an experience decision for the target data according to the end intelligent experience configuration, including: in response to the triggering of the target complex event, running the decision engine to match the timing of the embedding data according to the timing configuration information to determine whether the timing of the target complex event is met;

The contact engine in the operating end intelligent suite triggers a corresponding disposal plan according to the experience decision result, including: after the timing is successfully matched, running the contact engine to trigger the experience monitoring disposal plan corresponding to the target complex event.

In one implementation, after the unified event stream processing tool captures the tracking data, it also includes:

Perform client type and format analysis on the tracking data, and standardize the tracking data so that the data of different types of clients have a unified format and structure;

Among them, the standardized processing of the buried point data includes: parsing the buried point information, event ID, and parameter information of the buried point data; packaging the buried point information, event ID, and parameter information in a unified format, and converting them into event information in a standard format.

In one implementation, the decision engine is a complex event processing component based on a self-developed complex event matching rule library and declared as a static function, wherein the self-developed complex event matching rule library is developed to be compatible with multiple types of clients.

In one implementation, the terminal intelligent experience configuration includes an experience layering model;

The perception engine in the startup end intelligent kit obtains target data in real time, including: starting the perception engine to obtain static data of device hardware, real-time status data of the device, and real-time operation data of the scene;

The decision engine in the trigger end intelligent suite makes experience decisions for target data according to the end intelligent experience configuration, including: inputting the acquired static data of device hardware, real-time status data of device and real-time operation data of scene into the experience hierarchical model to predict the operation status of device;

The touchpoint engine in the operation-end intelligent suite triggers a corresponding disposal plan according to the experience decision result, including: the operation touchpoint engine triggers a corresponding experience layered disposal plan according to the prediction result of the device operation status.

In one implementation, the experience layering model is obtained by extracting features from data samples based on a convolutional neural network and inputting the extracted features into a deep neural network model for training.

In one implementation, the experience layering model is obtained by training a first network structure to obtain a first experience model and training a second network structure to obtain a second experience layering model. In the decision-making process of the decision engine, the first experience layering model is used as a white box model for white box pre-screening and the second experience layering model is used as a black box model for black box fine screening, so as to predict the operating status of the device.

In one implementation, the terminal intelligent experience configuration includes a full-link dynamic line prediction model;

The perception engine in the startup terminal intelligent suite acquires target data in real time, including: starting the perception engine to acquire user behavior data, user basic data, scenario data and performance data;

The decision engine in the trigger end intelligent suite makes experience decisions for the target data according to the end intelligent experience configuration, including: inputting the acquired user behavior data, user basic data, scenario data and performance data into the full-link dynamic line prediction model to predict the user's future behavior trajectory;

The touch engine in the operating end intelligent suite triggers a corresponding disposal plan according to the experience decision result, including: running the touch engine to trigger a corresponding experience enhancement disposal plan according to the prediction result of the user's future behavior trajectory.

According to another aspect of the present application, a terminal intelligent experience implementation device is provided, which is used to perceive, decide and handle user experience at the front end for front-end codes applicable to multiple types of clients, and the device is applied to a client of any type of client among the multiple types of clients, including:

A configuration request unit is used to start the client, send a terminal intelligent experience configuration request to the server, obtain the terminal intelligent experience configuration and store it locally;

The end intelligent execution unit is used to run the end intelligent suite in the front-end code, including: starting the perception engine in the end intelligent suite to obtain the target data in real time; triggering the decision engine in the end intelligent suite to make experience decisions for the target data based on the end intelligent experience configuration; and running the contact engine in the end intelligent suite to trigger the corresponding disposal plan according to the experience decision result.

According to one aspect of the present application, a storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the above-mentioned terminal intelligent experience implementation method when running.

According to one aspect of the present application, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the above-mentioned terminal intelligent experience implementation method.

With the help of the above technical scheme, the present application provides a method, device, medium and equipment for realizing end intelligent experience, which perceives, makes decisions and handles user experience at the front end for the front end code applicable to multiple types of clients. First, after starting the client, a request for end intelligent experience configuration is sent to the server to obtain the end intelligent experience configuration and store it locally; then the end intelligent suite in the front end code is run, including: starting the perception engine in the end intelligent suite to obtain the target data in real time; triggering the decision engine in the end intelligent suite to make experience decisions for the target data according to the end intelligent experience configuration; and running the contact engine in the end intelligent suite to trigger the corresponding handling plan according to the experience decision result.

The method for implementing the end intelligent experience provided in the embodiments of the present application emphasizes a cross-end, instant, and intelligent experience optimization method. It does not only focus on improving a certain aspect of performance, but also starts from the overall perspective of user experience, by real-time perception of user status and dynamic adjustment of experience to provide better services. Its core advantages include at least the following aspects:

1. Cross-terminal intelligent experience: Supports cross-terminal use, such as mini-programs, different technology stacks such as the web, etc. Among them, the front-end code is converted into multi-terminal support code through static compilation, which can provide consistent optimization effects on different platforms and devices;

2. Instant response capability: It can perceive user behavior and device status in real time, such as network conditions, memory usage, etc., and make instant adjustments based on this information. This instant response capability can continuously optimize the user experience during the user's use, rather than just a one-time optimization when the page is loaded;

3. Intelligent decision-making and automatic optimization: The end-to-end intelligent experience framework has built-in black box models and white box models, which can perform intelligent analysis based on the collected data to make targeted optimization decisions. This data-driven optimization method can more accurately identify and solve performance bottlenecks.

4. Global policy support: A series of global policies are provided, such as image quality degradation, package preloading, weak network switching, etc. These policies can be used flexibly in different scenarios to achieve end-to-end performance optimization.

Compared with cloud intelligence, edge intelligence has outstanding technical advantages:

1. Real-time performance: The real-time performance on the client is difficult to match on the cloud. For many interactive and streaming scenarios, the latency on the cloud can hardly meet the real-time experience requirements.

2. Offline capability: Although 4G and 5G are already popular, in real life, such as in cars, airplanes, or in the suburbs, network connections are often unstable or even unavailable. In this case, the intelligent terminal can effectively avoid the problem of poor network.

3. Data privacy: Due to data privacy requirements, a lot of user data cannot be uploaded to the cloud, so it can only be used on the client;

4. Cost: Although cloud computing provides massive computing power, the cost of AI computing and network bandwidth is very high for a large number of users.

In addition, compared with traditional performance optimization, traditional methods often focus on technical optimization for specific problems, such as code segmentation, lazy loading, resource compression, etc. The embodiment of the present invention is based on the end intelligent experience framework, and pays more attention to providing a comprehensive, dynamic and adaptive optimization solution from the perspective of user experience through real-time data analysis and intelligent decision-making. This method can not only solve some of the limitations encountered in traditional performance optimization, but also provide more personalized and high-quality services based on actual usage scenarios and user needs.

The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

BRIEF DESCRIPTION OF THE DRAWINGS

The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

FIG1 is a flow chart showing a method for implementing a terminal intelligent experience provided by an embodiment of the present application;

FIG2 is a schematic diagram of a terminal intelligent experience processing solution provided in an embodiment of the present application;

FIG3 shows a flow chart of a method for monitoring terminal intelligent experience provided in an embodiment of the present application;

FIG4 shows a schematic diagram of complex event configuration and matching logic in terminal intelligent experience monitoring provided by an embodiment of the present application;

FIG5 shows a flow chart of timing matching in terminal intelligent experience monitoring provided by an embodiment of the present application;

FIG6 shows a flow chart of a method for layering terminal intelligent experience provided in an embodiment of the present application;

FIG. 7 shows a schematic diagram of the logic of implementing the terminal intelligent experience layering provided in an embodiment of the present application;

FIG8 shows a flow chart of a method for enhancing terminal intelligent experience provided in an embodiment of the present application;

FIG9 is a flowchart of another method for implementing a terminal intelligent experience provided by an embodiment of the present application;

FIG10 shows a schematic diagram of the structure of a device for implementing a terminal intelligent experience provided in an embodiment of the present application.

Detailed ways

In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only embodiments of a part of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present application. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

All types of clients have user experience issues, so how to improve user experience is a topic that needs to be continuously studied in application development.

Traditional HTML 5 application development is efficient, but the user performance experience is not perfect, so there have been attempts to use Web technology to develop Web applications with user experience similar to native applications. The emergence of mini-programs (MiniApp) has quickly become popular with users. MiniApp is a small, installation-free, fast-loading program that usually runs in a host application or operating system (such as mini-programs, quick applications), or it can be a JS native application that supports cross-end deployment. Mini-programs themselves are limited by the architecture, and it is difficult to achieve a silky experience analogous to native applications. Especially for giant mini-programs, dynamic experience degradation during user operation is a difficult issue. Since the user experience problem of mini-programs is more prominent, the embodiment of this application takes mini-programs as an example for explanation, but it can be understood that it is applicable to other types of clients (native application APP, web pages, websites, etc.).

In existing solutions, the user experience is usually improved by collecting data on the front end, calculating the data on the server end, and then feeding it back to the front end. This method can be called server-side intelligence or cloud-side intelligence (cloud-side AI) solution. Although it has the characteristics of strong data processing capabilities, it has the limitations of high latency and high maintenance costs.

Therefore, the inventor of this application innovatively proposed to improve the user experience through end-side intelligence (or "terminal intelligence" or "front-end intelligence", hereinafter referred to as "end intelligence"), where end intelligence can be understood as the ability to think and make decisions on the end. In short, end intelligence is real-time perception, calculation, decision-making and intervention on the end side, which improves the user experience by enabling the end to have machine learning capabilities. The end-side intelligent experience in the embodiment of the present invention uses data analysis and machine learning technologies to deeply understand and understand user behavior, globally detect and coordinate various user device links and complex operating scenarios, and implements feedback loops and timely adjustments and interventions to optimize the client product experience. A comprehensive solution is characterized by immediacy, intelligence, and disposability.

After analysis and practice, end-to-end intelligence has outstanding technical advantages: 1. Real-time performance: The real-time performance on the end is difficult to match on the cloud. For many interactive and streaming scenarios, the delay on the cloud can hardly meet the real-time experience requirements; 2. Offline capability: Although 4G and 5G have been popularized, in real life, such as in cars, airplanes, or in the suburbs, the network connection is often unstable or even unavailable. At this time, end-to-end intelligence can well avoid the problem of poor network; 3. Data privacy: Due to data privacy requirements, many user data cannot be uploaded to the cloud, so they can only be used on the end; 4. Cost: Although cloud computing provides massive computing power, the cost of AI computing and network bandwidth is very high for massive users. Now that end-side devices are popular, there are a large number of mobile phones, smart cars, set-top boxes and other devices with computing power that can be used to reduce the cost of cloud deployment.

The inventors of this application continued to study and found that there are several challenges in realizing end intelligence. The first is the challenge of package size. The process technology on the end involves occupying the installation package size and storage resources of the APP, which puts forward high requirements on the end computing engine slimming, end model size, end model compression, dynamic publishing and other capabilities; the second is the performance challenge. Due to the use of computing resources of mobile devices, it is necessary to balance user experience and computing efficiency. It is necessary to continuously optimize the mobile device CPU/GPU, memory, power consumption, instruction set, thread scheduling, etc.; the third is the adaptation challenge. In the face of the fragmentation of devices and system versions, end intelligent technology needs to be adapted in large quantities to ensure stability and availability during operation. Therefore, the embodiment of this application proposes a solution for implementing end intelligent experience, which aims to overcome the defects of cloud AI and optimize the implementation effect of the solution based on the above challenges.

In addition, in existing network applications, the same network product often corresponds to different types of clients (published through different channels). For example, a takeaway product, in addition to the takeaway APP, also has various types of client forms such as mini-programs, websites, and web pages. For the same product front-end code, if different terminal intelligent experience mechanisms are adopted for various types of clients (including but not limited to APP, mini-programs, websites, web pages, etc.), it will have a great impact on code development and publishing efficiency. Therefore, the embodiment of the present application is also further improved in this regard, in order to achieve a set of terminal intelligent experience implementation solutions applicable across terminals. In order to achieve a solution applicable to multiple types of clients, the embodiment of the present application relies on a cross-terminal programming language (such as JavaScript, JS) for front-end code development. Since JS development is compatible with APP, mini-programs, and H5 clients of various types, the purpose of cross-terminal can be achieved. In the specific implementation process, it is also necessary to consider processing the source code to remove the differences between various types of clients.

Refer to Figure 1, which is a flow chart of a method for implementing a terminal intelligent experience in an embodiment of the present application. The method is used to perceive, decide and handle user experience at the front end for front-end codes applicable to multiple types of clients. The method is applied to a client of any type of client among multiple types of clients, and includes the following steps S101-S102.

In the embodiment of the present application, the front-end code (e.g., JavaScript (JS) front-end code) applicable to multiple types of clients (also called "cross-client") is obtained by statically compiling the source code to remove the differences between various types of clients. Static compilation means that when the compiler compiles the executable file, it extracts the part of the corresponding static library that the executable file needs to call and links it into the executable file, so that the executable file does not rely on the dynamic link library when running. In the embodiment of the present application, static compilation of source code is mainly to process the parts of JS source code that are strongly constrained and cannot be dynamically modified, which may include three aspects: 1. Module reference, such as replacing module references and modifying suffix names in JS source code; 2. Template attribute mapping or syntax compatibility processing, for example, in JS, template attribute mapping generally refers to mapping the attributes of an object to a template string to generate a complete string. This mapping can be used to create dynamic HTML strings or other types of text; in JS, if the purpose of syntax compatibility processing is to ensure that the source code can run in different environments, it is necessary to pay attention to the syntax differences that may exist in different environments. The syntax compatibility processing of the scene can be carried out in a variety of ways, for example, using strict mode, applying JSON objects to serialize and deserialize, using optional chain operators, using specific syntax, using template import or export, etc.; 3. Configuration mapping, such as through page configuration. Therefore, through the above-mentioned operations, the differences between various types of clients are smoothed out, providing a basis for realizing cross-end intelligent experience.

S101: Start the client, send a request for terminal intelligent experience configuration to the server, obtain the terminal intelligent experience configuration and store it locally;

S102: Running the end intelligent suite in the front-end code, including: starting the perception engine in the end intelligent suite to obtain target data in real time; triggering the decision engine in the end intelligent suite to make experience decisions for the target data based on the end intelligent experience configuration; and running the contact engine in the end intelligent suite to trigger corresponding disposal plans according to the experience decision results.

In an embodiment of the present application, in order to reduce the computing processing pressure on the end, the end intelligent experience configuration is performed in advance on the server side. After the client is started, the end intelligent experience configuration is loaded and stored locally to prepare for subsequent local decision-making and other operations.

The end-to-end intelligent experience configuration pre-configured on the server (or cloud) can be understood as including two dimensions of configuration. 1. Event rule configuration: The server needs to pre-define various user behavior events and related rules, including event triggering conditions, rule execution logic, etc. 2. Timing strategy configuration: In the scenario, timing refers to the expression of performing a certain specific operation when a series of conditions are met. In implementation, timing is a complex event (compound event) rule plus a trigger configuration; the strategy is the condition that needs to be met for the trigger timing or the compound event rule model deployment: For strategies that require machine learning model support, the server is responsible for the deployment and distribution of the model. These models are trained based on a large amount of historical data and are used to predict user behavior or evaluate the quality of user experience.

In the client code development, an end-to-end intelligent kit is embedded. The end-to-end intelligent kit can be understood as a set of functional modules that solve the end-to-end intelligent experience. It has a deep insight into and understanding of user behavior, combines data analysis, machine learning and other technologies, comprehensively detects and coordinates the user's device environment and scenario operation scenarios, and optimizes the client experience through real-time feedback loops and timely adjustment interventions. The end-to-end intelligent kit has built-in perception engines, decision engines and contact engines. According to the end-to-end intelligent experience configuration loaded locally, the corresponding engine is automatically triggered at the corresponding processing node to achieve real-time perception, real-time decision-making and real-time processing of data.

In one implementation, an end-to-end intelligent suite can be developed based on the end-to-end intelligent experience architecture. The end-to-end intelligent experience framework may include the following parts. 1. End-to-end intelligent experience solution layer, a set of solution strategies and tools for specific problems, which rely on the monitoring and decision-making capabilities provided by the basic capabilities to solve various problems encountered by users during use. According to the experience dimension classification, the end-to-end intelligent experience solution layer may include three functional modules: experience monitoring, experience stratification, and experience enhancement.

2. End-to-end intelligent experience plug-in system, such as a plug-in market and plug-in development platform, allowing third-party or internal developers to develop plug-ins for specific functions. This system can be seen as an extension platform, relying on the services and APIs provided by the infrastructure and solutions to develop and distribute plug-ins.

3. The end-to-end intelligent experience core is the basis for the operation of the entire system. It can be composed of three parts: perception, modeling, decision-making and disposal. It is responsible for collecting and processing information and making intelligent responses. This core directly supports the intelligent implementation of experience monitoring, experience enhancement and experience stratification.

4. The cross-end intelligent layer is the high-level functional layer of the system, which integrates capabilities such as unified event stream, model reasoning, feature engineering, decision engine, and touchpoint engine.

The logical relationship between the above parts and levels reflects the service hierarchy from basic to advanced. Each layer provides support and services to the upper layer, and they work together to ensure the overall intelligent and efficient operation of the system.

It can be seen that the above-mentioned end-to-end intelligent suite developed through the end-to-end intelligent experience architecture can be understood as being built on a unified experience base. Unlike scattered solutions (such as developing separate specific solutions for specific experience problems), the end-to-end intelligent experience implementation solution is modeled on a unified experience base. This approach of using a unified experience infrastructure has the benefit of reducing costs and increasing efficiency. At the scene level, it only needs to call the base and render to meet the personalized needs of the scene, which has the advantages of scalability and personalization.

The following is a detailed description of the three main actions in S102 (data collection, decision-making, and disposal).

1. Data Collection

The end intelligence suite includes a perception engine for collecting data. Perceiving the real-time data of users on the end is the basis of the end intelligence experience. The target data in S102 is a collection of different types of data that need to be perceived in different experience solutions (for example, including experience monitoring, experience layering, and experience enhancement, which will be introduced in detail later). Generally speaking, the target data may include user basic data, user behavior data, scenario data, performance data, etc.

User basic data can be obtained through some APIs, such as using getSysteminfo to obtain the device model, using getNetworkType to obtain the current network type, and using onNetworkStatusChange to monitor network status change events.

User behavior data can be tracked through UT and placed in different event streams for tracking and reporting according to the event number to which the behavior belongs. User behavior data is relatively complex and requires certain calculation logic. Taking the "page dwell time event" as an example, the following points are involved: (1) The timing starts when the user enters the page, and the page dwell event is triggered after the user browses the page for more than xx seconds; (2) When the user stays on the page for less than the required time, that is, jumps out of the page, the dwell time will be paused, and the accumulation will continue after returning to the page; (3) The page dwell event is triggered at most once in a page visit (pageAppear to pageDisAppear).

For scenario data, it may involve behaviors such as page visits (2001), clicks (2101), and exposures (2201). Other behaviors are placed in custom behaviors, and then different scenario data are sorted out according to certain scenario rules.

For performance data, such as interface time consumption, it can be collected and reported through the basic network SDK. Rendering-related performance data can rely on the client framework and container (such as the mini-program framework and container), combined with the scene/page rendering scene, to complete the data collection and reporting of each stage. Taking the mini-program as an example, it is specifically divided into: (1) Mini-program performance plug-in: collect application/page startup and standard life cycle time consumption, and report automatically, among which, (1.1) expansion capability: through the mini-program performanceApi, provide some container startup performance data, mainly the startup starting time and route time consumption; (1.2) for some channels lack container collection API, unable to obtain the startup time point, you can customize the scene T2 time consumption with the application onLaunch as the starting point; (2) Mini-program performance component mini-performance collects the scene T2 time consumption end point, among which, (2.1) settlement method: the component is based on the IntersectionObserver API, and is connected to the "last module of the first screen full rendering" by the scene. The component will record the event (& time) of being seen (& rendered) in the view as the settlement point of simulated T2, and there is no additional performance loss. In addition, the Mini Program framework provides the onError method of the App object, which can be used to monitor global script errors, and onPromiseError to monitor all uncaught Promise exceptions. As for the monitoring of page white screen, a custom method can be adopted, which starts a timer after the interface request is completed and the data is returned. If no element on the page is successfully rendered within the next few seconds, this situation is regarded as a page white screen and corresponding statistics are performed.

Decision-making

The end intelligence suite includes a decision engine for making experience decisions, that is, making experience decisions based on the downloaded end intelligence experience configuration and the target data collected in real time. Among them, the end intelligence experience configuration may include event configuration information or model configuration information, so as to match events according to the event configuration information, or make predictions according to the model. Among them, the event configuration information can be, for example, complex event configuration information (corresponding to the "experience monitoring" solution, which will be described in detail later), and the model can be an experience layering model (corresponding to the "experience layering" solution, which will be described in detail later), or a full-link dynamic line prediction model (corresponding to the "experience enhancement" solution, which will be described in detail later). In an embodiment of the present invention, event configuration and model training and deployment are offline, and reasoning is performed at the front end. The front end can achieve faster response speed, reduce network request latency, protect user data privacy, and still provide services in an environment without network connection.

3. Disposal

The end intelligent kit includes a contact engine, which is used to trigger the experience disposal according to the experience decision result. In the end intelligent experience solution provided by the embodiment of the present application, after the problem is perceived by collecting and experiencing the decision, the problem can be solved or reported directly on the end, and a comprehensive and flexible disposal capability can be provided. For example, in some cases, it is necessary to interact with the user through UI forms such as pop-ups and bottom floating bars, and real-time messages can be pushed on the application interface to inform the user of important information or operation feedback, such as real-time message prompts, user guidance, etc. The end intelligent experience framework can provide a complete UI component, which is highly customizable. The component is built into the host in the form of dynamic plug-ins/sub-packages, and the information can be quickly reached to the user. The end intelligent experience framework can intercept when sending or receiving network request data, customize the request header or process the returned data, so as to flexibly intervene in the network. The end intelligent experience framework can dynamically proxy and rewrite the basic library API in different scenarios to expand or modify the original behavior. For example, the hashToUrl change can be proxied to return pictures of different qualities under different operating states.

Therefore, through the above three steps (data collection, decision-making, and processing), real-time perception and processing of user experience can be performed on the terminal. The following is an example of the above steps using a specific example of "experience layering".

Step 1: Preparation steps and model training. Data collection: Collect behavioral data of users visiting pages, including page loading time, user device type, network environment and other information. Feature extraction: Extract key features from the collected data, such as device performance indicators (CPU, memory usage), network speed, etc. Model training: Use these features and page loading time as training data to build a machine learning model to predict the operating status of different models. Model publishing: Convert the trained model to ONNX format and publish the model file to CDN for the decision engine to load and predict.

Step 2: Real-time feature collection. Data collection: When users use the app, real-time monitoring and collection of data such as page loading time, device CPU usage, and memory usage. Data normalization: Adjust the data value range to a specific interval, usually between 0 and 1 or -1 to 1.

Step 3: Use the model to make decisions. The collected feature data is used to determine the model's operating status based on the model's prediction results. If the prediction result exceeds the set threshold, it is determined to be in a low operating state (poor operating state or average operating state).

Step 4: Scenario handling. Interface loading optimization: For models that are judged to be in poor operating status, automatically optimize the network loading strategy, such as delaying the loading of non-critical interfaces; Content adaptation: Dynamically adjust the page content according to the device operating status, such as reducing the image quality or simplifying the page elements; Data feedback: Continue to collect handling results and user experience feedback as new data input to continuously optimize models and strategies.

As described above, from the perspective of user experience, it can include three dimensions: experience monitoring, experience stratification, and experience enhancement. See Figure 2, which is a schematic diagram of the end intelligent experience solution provided by the embodiment of the present application. Experience monitoring is to actively discover the user's experience problems (such as whether a specific complex event has occurred) in the process of using the client through real-time collection, analysis and early warning of user experience data, and provide real-time feedback and on-site disposal, thereby improving the user experience. Experience stratification is to evaluate the operating status of the user's device from the perspective of actual user use. Different from the rigid classification of devices based on factory hardware information in the past, the intelligent experience stratification is based on the multi-dimensional time-consuming data of the actual operation of the application, and uses the experience stratification model (obtained through the end intelligent experience configuration) to comprehensively and weightedly score the user's device data in real time, and perform corresponding disposal according to the score. Experience enhancement: Based on the user's real-time interactive behavior information, the full-link dynamic line prediction model (obtained through the end intelligent experience configuration) is used to realize real-time user future trajectory estimation, preload the sub-packages and pages that the user may visit, and enhance the user experience.

Therefore, in one implementation, the above-mentioned processing scheme includes: an experience monitoring processing scheme, an experience layering processing scheme, and/or an experience enhancement processing scheme.

The experience monitoring and handling scheme includes: when a complex event is determined to occur according to the experience decision, real-time message push and/or page jump operation are performed in the interface in a UI interactive manner.

The experience layering disposal scheme includes: when it is determined according to the experience decision that the device operating state is a low operating state, an interface loading optimization and/or page content dynamic adjustment scheme is performed. Among them, the prediction results of the experience decision can be divided into three levels: good operating state, general operating state, and poor operating state. Among them, good operating state belongs to a high operating state, and general operating state and poor operating state belong to a low operating state. Among them, the general operating state can be understood as a medium level of operating state, with room for experience improvement, and the poor operating state is in urgent need of experience improvement. Those skilled in the art can set and distinguish the operating states of each level according to specific performance parameters. When determining a low operating state, a corresponding disposal scheme can be executed, for example, improvement can be made through an interface loading optimization scheme, including optimizing network loading strategies and the like; for example, a page content dynamic adjustment scheme may include: reducing image quality, simplifying page elements, and/or simplifying page interactions.

The experience enhancement solution includes: obtaining a user's future behavior trajectory based on the experience decision, determining sub-package resources or page resources that the user may access, and preloading the sub-package resources or page resources.

In one implementation, the execution logic relationship between different treatment schemes can be determined by setting their priorities. Therefore, the above method also includes: setting different priorities corresponding to different treatment schemes, and determining the execution logic relationship between different treatment schemes according to the priorities. For example, the experience layering treatment scheme is set to correspond to the first priority; according to the processing result of the experience layering treatment scheme, determine whether to execute the experience monitoring treatment scheme and/or the experience enhancement treatment scheme. The advantage of this processing method is that it is first processed through the experience layering treatment scheme, such as reducing the image quality for low-performance devices. After the above-mentioned experience layering treatment, if the device operating state is still in a "poor" state (which can be predicted again by the layered experience model), then it is considered that it is not suitable for experience monitoring and experience enhancement at this time, so as to save resource overhead and reduce data processing pressure.

In addition, in one implementation, after executing the treatment plan, the following may be included: determining the treatment plan to be executed and collecting user feedback data, sending the treatment plan and user feedback data to the server as new inputs to enable the server to optimize the terminal intelligent experience configuration. Thus, the treatment results and user experience feedback are continuously collected as new data inputs to continuously optimize the model and strategy.

In addition, in order to optimize the implementation effect, decision-making is performed through the white box + black box model.

White box model: When running online, the weights are already specific, and there is no need to load the inference model for inference. The white box model is mainly used to screen low-performance devices in the cold start phase when the user opens the client (such as a mini program) to avoid the model running bringing additional burden to the device. This strategy ensures that further operation status diagnosis and experience optimization can be performed through more advanced models in subsequent processes. Black box model: This is a deep network architecture constructed by methods such as multi-layer nonlinear transformation and abstract global feature learning, which simulates the hierarchical processing of information by the human brain nervous system and is used to solve classification, regression and other problems. For example, in an embodiment of the present application, the full-link dynamic line prediction model can mainly rely on a black box model, such as a deep learning network (such as LSTM), to predict the user's behavior path. This model analyzes the user's historical behavior data and attempts to predict the user's next possible operation or preference. Because deep learning models usually involve complex network structures and a large number of parameters, their internal decision-making logic is opaque to users, so they are called black box models. This type of model plays an important role in predicting users' future behavior trajectories and realizing intelligent preloading, aiming to improve the fluency and personalization of user experience. The experience layering model uses more white-box models, such as the GBDT algorithm, to manage the user experience in layers according to the device's operating status. By analyzing the device's performance parameters (such as memory usage, processor performance, etc.), the model can evaluate the device's performance level in real time and adjust the application's performance accordingly, such as reducing image quality or simplifying interaction effects to adapt to devices of different performance levels. The decision-making process of the white-box model is transparent, allowing developers to clearly understand the model's decision logic and basis, which is convenient for tuning and maintenance.

Therefore, in one implementation, experience decisions for experience monitoring, experience stratification, or experience enhancement are made through a white box model or a black box model, and preliminary decisions are made based on the white box model, and it is determined whether to use the black box model to make further refined decisions based on the preliminary decision results of the white box model.

In one implementation, the experience layering model is set to a white box model mode; in the cold start phase of the client, the experience layering model in the white box model mode is run, and according to the prediction results of the experience layering model, it is determined whether to continue to execute the end intelligent suite or shut down the end intelligent suite; accordingly, the full-link traffic flow prediction model is set to a black box model mode; in the hot start phase of the client, the full-link traffic flow prediction model in the black box model mode is run.

It can be seen that the end intelligent experience implementation method provided by the embodiment of the present application emphasizes a cross-end, instant, and intelligent experience optimization method. It does not only focus on improving a certain aspect of performance, but also starts from the overall perspective of user experience, by real-time perception of user status and dynamic adjustment of experience to provide better services. Its core advantages may include:

1. Cross-terminal intelligent experience: Supports cross-terminal use, such as mini-programs, different technology stacks such as the web, etc. Among them, the front-end code is converted into multi-terminal support code through static compilation, which can provide consistent optimization effects on different platforms and devices;

2. Instant response capability: It can perceive user behavior and device status in real time, such as network conditions, memory usage, etc., and make instant adjustments based on this information. This instant response capability can continuously optimize the user experience during the user's use, rather than just a one-time optimization when the page is loaded;

3. Intelligent decision-making and automatic optimization: The end-to-end intelligent experience framework has built-in black box models and white box models, which can perform intelligent analysis based on the collected data to make targeted optimization decisions. This data-driven optimization method can more accurately identify and solve performance bottlenecks.

4. Global policy support: A series of global policies are provided, such as image quality degradation, package preloading, weak network switching, etc. These policies can be used flexibly in different scenarios to achieve end-to-end performance optimization.

In summary, compared with traditional performance optimization, traditional methods often focus on technical optimization for specific problems, such as code segmentation, lazy loading, resource compression, etc. The embodiment of the present invention is based on the end intelligent experience framework, and pays more attention to providing a comprehensive, dynamic and adaptive optimization solution from the perspective of user experience through real-time data analysis and intelligent decision-making. This method can not only solve some limitations encountered in traditional performance optimization, but also provide more personalized and high-quality services based on actual usage scenarios and user needs.

The following is an illustrative description of the terminal intelligent experience implementation method provided in the embodiment of the present application from three perspectives: experience monitoring, experience stratification, and experience enhancement.

First, experience monitoring is described with reference to FIG. 3 to FIG. 5 .

Experience monitoring is to proactively discover user experience issues (such as whether specific complex events have occurred) when using the client through real-time collection, analysis and early warning of user experience data, and provide real-time feedback and on-site disposal to improve user experience.

Referring to FIG3 , a flow chart of a method for monitoring terminal intelligent experience provided in an embodiment of the present application is shown.

S301: Start the client, send a terminal intelligent experience configuration request to the server, obtain complex event configuration information and initialize it, where the complex event configuration information includes event configuration information and timing configuration information of at least one target complex event;

S302: Start the perception engine, capture the embedded point data based on the unified event stream processing tool in the front-end code, and determine whether the target complex event corresponding to the event configuration information is triggered according to the embedded point data;

S303: In response to the triggering of the target complex event, the decision engine is run to match the timing of the buried point data according to the timing configuration information to determine whether the timing of the target complex event is met;

S304: After the timing is successfully matched, the contact engine is run to trigger the experience monitoring and handling solution corresponding to the target complex event.

Complex events can be understood as combined events. For example, in the actual application of big data analysis, it may be necessary to detect the occurrence of "continuous login failure events", which is actually a combination of "login failure" and "login failure". Another example is that it may be necessary to detect the user's order payment behavior, which is also a combined event. The "payment event" occurs within a period of time after the "order event", which also includes time restrictions. Combined events like the above multiple events are called complex events.

In an embodiment of the present application, complex event configuration information can be configured in advance or in real time through an end intelligent configuration platform, which can be a visualization platform and can be set on the server. Through the end intelligent configuration platform, users (such as R&D personnel) can configure complex event information. For example, the end intelligent configuration platform provides configuration of timing decisions, including event configuration and timing configuration. Event configuration refers to the matching rules of user events on the end, such as using "spm== a.b.c&&click" to correspond to a click event at a fixed position. The timing configuration is a sequence of events, which is used to perform complex matching on the occurrence sequence of user event streams on the end, such as "click to enter the store" then "add to cart" and then "return to homepage". In particular, in order to perform personalized configuration for different types of clients, the corresponding complex event set can be configured for different types of clients through the end intelligent configuration platform, wherein the complex event set includes multiple complex events, and each complex event includes event configuration information and timing configuration information.

See Figure 4, which is a schematic diagram of complex event configuration and matching logic in end-to-end intelligent experience monitoring. In this example, complex event configuration is first performed on the configuration management platform, including event configuration and timing configuration, where event configuration (which can be understood as feature configuration) refers to the accumulation and statistics of user behaviors, such as the number of times a store is entered within 3 days; timing configuration refers to specific conditions that can trigger decisions and responses, such as entering a store twice and adding purchases. After the configuration management platform is configured, the configuration information can be written into a database DB, and then a configuration gateway service reads data from the DB. After the client is started, the client automatically loads the configuration information from the configuration gateway service. After obtaining the configuration information, the end-to-end intelligent suite of the client makes relevant decisions based on the configuration information. Specifically, the decision engine performs event mapping/filtering, pattern matching, behavior accumulation, etc. to complete the timing decision. After the decision engine determines that the timing match is successful, it runs the contact engine to trigger the corresponding disposal plan, for example, real-time message push in the interface through UI interaction, or automatic page jump operation for the current operation failure event (for example, front-end cashier jump for multiple payment failures).

When the client is started, the complex event configuration information will be automatically loaded. For example, in step S301, by sending a request to the server, the complex event configuration information is obtained and initialized, wherein the complex event configuration information includes at least one event configuration information and timing configuration information of a target complex event. The complex event configuration request may carry latitude and longitude information, terminal information, and user ID, wherein the terminal information may refer to the client type information. After receiving the complex event configuration request, the server may determine the client type according to the terminal information, thereby matching the complex event configuration information corresponding to the type of client and issuing it.

In the embodiment of the present application, the unified event stream processing tool refers to an event stream standardization processing tool based on super location tracking, which can classify user behaviors into standard events based on tracking information of various types of clients (APP, mini-programs, H5 pages, etc.), including tracking information, event ID (eventId), parameter information (properties) and other information fields. Functions in other pages or plug-ins can obtain the operations performed by the current user on the terminal by listening to the unified event.

Among them, super location tracking includes the following settings.

For example, event marking is used to mark an event in a page through a.b.c.d, where a represents the user-side level, b represents a page within the terminal, c represents a block in a page, and d represents a specific click position in a block.

Event type The abcd bits above can only mark a specific location, and different event types are distinguished. Different general parameter information is recorded based on specific event types to cover all statistical needs. It also helps to reduce the difficulty of data processing and resource consumption. For example, the following three types can be used: Page type: In order to count user behavior at the page granularity, when entering and exiting the page, it is necessary to record parameters such as user id, device id, timestamp, entry or exit, etc., so that the UV, PV, user stay time and other data of each page can be counted; Click type: Long press, sliding, etc. are all counted as click types; Exposure type: refers to the exposure at a specific module level, such as product exposure and video exposure. By analyzing click-type events together, the click rate of a product, post, or video can be calculated.

In terms of jump relationships, it is sufficient to record the relationship between the previous jump and the previous two jumps in the URL. For example, for "clicking the first product on the shopping cart page", a click event can be reported: but the previous jump may be from a product detail, in which case the previous jump can be recorded; the previous two jumps may be from a product in the "Guess You Like" information flow on the homepage, in which case the previous two jumps can be recorded. Based on this, the user's movement lines are connected in series.

Traffic channel For example, if a channel on the homepage contains multiple pages, traffic channels are introduced. For example, multiple pages contained in a flash sale channel are marked as a traffic channel. Then, as long as the number of visitors to these three pages is deduplicated, the total UV of this channel can be obtained. Other clicks, total page visits, etc. can also be processed in the same statistical logic.

Among them, after the unified event stream processing tool captures the buried point data in the above step S302, it can also include: performing client type analysis and format analysis on the buried point data, and standardizing the buried point data so that the data of different types of clients have a unified format and structure, wherein the standardized processing of the buried point data can specifically include: parsing the buried point information, event ID, and parameter information of the buried point data; encapsulating the buried point information, event ID and parameter information in a unified format, and converting them into event information in a standard format.

It can be seen that the unified event stream processing tool in the embodiment of the present application can be understood as a standardized event stream processing tool for the C-end, which uses the super-position buried points in the cross-end (APP, mini-program or H5) source code to capture the buried point data, and encapsulates the buried point data and then converts it into standardized events. By introducing a unified event stream processing tool, the target complex events triggered by the monitoring processing tool are monitored. When the target complex event occurs, the filtering and matching of the complex event timing decision engine is triggered. This unified event stream processing method is an important part of realizing cross-end processing of complex events. By capturing buried point data based on super-position buried points, event information of various types and locations can be captured on the C-end, and by standardizing the buried point data, it makes it possible to use a unified complex event timing decision engine to detect and handle complex events based on the same set of complex event matching rule bases and the same matching logic.

In an embodiment of the present application, a decision engine is used to make timing decisions for complex events (which may be referred to as a "timing decision engine"), which may be based on a self-developed complex event matching rule library and declared as a complex event processing component of a static function, wherein the self-developed complex event matching rule library may be developed using JavaScript to be compatible with multiple types of clients. The advantage of the self-developed complex event matching rule library is that it can be personalized or modified according to scenario requirements, and specific personalized rule matching is performed based on this complex event matching rule library. Among them, in order to implement a cross-end adaptive timing decision engine, the complex event processing component can be declared as a static function through the aforementioned JS static compilation processing method to remove the differences between multiple ends and obtain a cross-end timing decision engine.

In one implementation, the specific processing process of the complex event timing decision engine is: according to the pre-configured complex event matching rules, the target complex event is subjected to event processing and timing matching, wherein the complex event matching rules include event rules, pattern rules and timing rules. For example, for the triggered target complex event, event filtering is performed according to the event rules, and event conversion is performed on the filtered events to obtain event IDs in a standard format; pattern filtering is performed on the converted events according to the pattern rules, and the events corresponding to the timing rules are retained, and pattern matching is performed on the retained events, wherein rule matching is performed according to multiple rule rules and item rules corresponding to the pattern rules to determine whether a round of pattern matching is successful, and whether the pattern matching is successful is determined according to the progress of multiple rounds of pattern matching; after the pattern matching is successful, fatigue detection is performed according to the timing rules, wherein the latest fatigue detection is performed according to the fatigue cache information, and if the fatigue detection passes, it is determined that the timing matching is successful.

Among them, pattern matching in complex event processing can be understood as two aspects: the characteristics of each simple event and the combination relationship between simple events. In addition, the function of the pattern can be expanded, such as the time limit for matching detection, whether each simple event can be repeated, and whether to skip the subsequent matches after encountering a match for the pattern of event recurrence. Pattern matching detects continuous events or discontinuous but sequential events based on the neighbor relationship and conditions in each event. The pattern may also have a time limit. If the matching conditions are not met within the set time range, the pattern matching will time out.

Referring to FIG. 5 , an example of the timing decision engine processing is shown.

S501, event triggering: monitoring a target complex event triggered by a unified event stream processing tool.

S502, event filtering: filtering according to event rules. For example, if only click events are configured currently, all exposure events and scrolling events will be filtered to reduce the subsequent matching pressure.

S503, event conversion: converting the filtered events into unified events that can be matched, wherein the front-end events are mapped to the IDs of the events configured in the back-end.

S504, pattern filtering: There are multiple complex rules in one opportunity, and the events configured for each rule are different. In order to reduce the matching pressure, further filtering is performed here to keep only the events used by this rule.

S505, pattern matching: for example, by traversing the rules in depth first, the events are matched in sequence, including multiple rule rules. Among them, pattern matching further includes rule matching and item matching, wherein the rule rule refers to a rule in a complex rule, including multiple items; item matching: the item corresponds to the event configured in the background one by one, and the underlying event matching is performed here, and the matching results are returned by level after completion.

S506, check the matching progress: the progress is increased by one each time a pattern match is completed, and the next step is entered after all are completed.

S507, matching result: determining whether the matching is successful or failed.

S508 Fatigue detection: After the match is successful, fatigue is detected. For example, the rule is set as follows: trigger three orders within 3 days; if the order is less than or equal to 3 times within 3 days, fatigue is true (true), and subsequent callback behavior is allowed; if the order is more than 3 times within 3 days, fatigue is false (false), and subsequent callback behavior is no longer allowed. Among them, the historical fatigue can be cached locally and loaded next time the client is entered.

Step 509: Decision completed.

In one implementation, after the timing match is successful, the callback function (e.g., callback function) pre-registered by the complex event timing decision engine can be automatically called, and the contact engine can be started through the callback function to execute the disposal plan corresponding to the target complex event. Wherein, as described above, the terminal intelligent configuration platform can realize the configuration of complex events. Correspondingly, the terminal intelligent configuration platform can receive the timing match success information of the target complex event, and determine the corresponding disposal plan according to the scene attributes of the target complex event and/or the terminal attributes of the corresponding type of client of the target complex event. Wherein, the scene attributes can include the application scene type information, and the client terminal attributes are used to distinguish the client type. Therefore, the above method can be used to realize the disposal of personalized disposal plans for the target complex event, wherein the disposal plan can be, for example, a prompt window or a floating layer.

Taking "user automatically switches to the front checkout counter when payment fails three times in a row" as an example, experience monitoring is explained illustratively.

Step 1: On the client intelligent configuration platform, configure the user payment failure event (if it has already been configured, it can be reused). Here, it is assumed that the event is event F.

Step 2: On the end intelligent configuration platform, configure the number of times the user payment failure event occurs to three times and the number of times event D matches to 3;

Step 3: After starting the client, the C end requests configuration information and performs initialization;

Step 4: After the user fails to pay once (determined by the unified event stream processing tool through tracking), the timing decision engine is triggered to match the timing, and the number of times is recorded as 1;

Step 5: After the user fails to pay twice, the timing decision engine is triggered to perform timing matching, and the number of times is recorded as 2;

Step 6: After the user fails to pay three times, the timing decision engine is triggered to perform timing matching, and the number of times is recorded as 3;

Step 7: After the matching test is passed, the registered callback function (such as callback function) is called;

Step 8: Run the touchpoint engine and automatically switch to the front cashier. The front cashier can select a specific payment platform on the merchant page to complete the payment, avoiding jumping to the intermediate page and helping to optimize the online banking payment experience. The front cashier supports integration on the Web, WAP, and App terminals.

In the above example of "user failed to log in three times in a row", the timing decision engine is an engine that matches according to the matching rules. It only performs data parsing and rule matching. It can be understood as a core component that is declared as a static function but caches part of the execution process. The association between matching rules and complex events can be pre-configured on the server side, while the intermediate links in the execution process (triggering complex events and timing matching) are not associated, which can be understood as a black box implementation method. Taking "no response after repeated clicks at the same location on the monitoring page" as an example, the monitoring strategy is first configured. For example, the monitoring strategy is configured according to the needs of the scenario, and the matching rules and triggering timing are agreed upon: event matching logic: distinguish pages by ab positions, distinguish scenarios by c positions, and distinguish pit positions by d positions. All clickable areas on the page can be covered by a timing configuration; trigger strategy: 3 or more consecutive clicks within 2s; then, user anomalies are reported in real time through embedding points, and anomalies that can be handled are intervened in real time. For example, targeted processing is made for different points, such as page downgrade, jump to H5 page details page, etc. Therefore, compared with traditional monitoring that is difficult to perceive occasional anomalies, the experience monitoring in the embodiment of the present application can discover problems in the first time based on user sessions, and intervene in processing before customer complaints are generated, thereby improving user experience.

In addition, taking "intelligent correction of bill of lading mobile phone number" as an example, in the scenario where the user places an order, a decision is made based on the similarity score distribution. When it is determined that the user is likely to have filled in the wrong mobile phone number (for example, the score is between 0.7-1), the mobile phone number is corrected to remind the user to modify the mobile phone number with one click (this can be achieved on the front end). For example, correction is performed through bubble prompts, and the mobile phone number that the user will fill in can be predicted and the one-click filling function can be provided.

In addition, there are many other scenario instances of experience monitoring, such as automatic switching of mobile networks in weak WIF scenarios, etc., which are not described one by one in the embodiments of this application.

In one implementation, the method further includes the following steps: determining whether the target complex event is a real-time behavior event or a cumulative behavior event; if it is a real-time behavior event, obtaining user behavior data in real time, and performing complex event trigger judgment and timing matching based on the buried data corresponding to the user behavior data in real time; if it is a cumulative behavior timing, obtaining user behavior data in multiple times, accumulating the user behavior data, and performing complex event trigger judgment and timing matching based on the buried data corresponding to the currently accumulated user behavior data. It can be seen that through the above method, using user behavior data as the basis for matching can achieve more intelligent decision-making. Unlike processing real-time user behavior events for users, processing cumulative behavior events (for example: ordering 5 times in a week) can achieve domain linkage: accumulated user data can be used in the process of timing decision-making, and a complete access solution can be provided.

It can be seen that through the above-mentioned experience monitoring technical solution, the embodiment of the present application is for the JavaScript front-end code applicable to multiple types of clients, and complex events are detected and handled at the front end, wherein after the client is started, a complex event configuration request is sent to the server, complex event configuration information is obtained and initialized, and then based on the unified event stream processing tool in the front-end code, the buried data is captured, and it is determined whether the target complex event corresponding to the event configuration information is triggered according to the buried data; then, in response to the triggering of the target complex event, the decision engine is run to match the timing for the buried data according to the timing configuration information to determine whether the timing of the target complex event is met; finally, after the timing is matched successfully, the callback function is triggered to execute the handling plan corresponding to the target complex event. In the embodiment of the present application, compared with the method of mainly relying on the server to detect and handle complex events, complex event processing is performed on the C-end side, thereby real-time rule matching can be performed for the real-time acquired data, thereby reducing the processing time, and having the advantages of low latency and low maintenance cost.

Moreover, unlike the existing complex event processing scheme that can only be applied to one type of client, the embodiment of the present application also makes improvements in this regard, in order to achieve a set of complex event processing schemes applicable across terminals. In order to achieve a complex event processing scheme applicable to multiple types of clients, the embodiment of the present application relies on JavaScript (JS) for front-end code development. Since JS development is compatible with APP, mini-programs, and H5 clients, it can achieve the purpose of cross-terminal. Through a set of event processing mechanisms, it can be universal for various types of clients, greatly reducing the development pressure.

In addition, in order to facilitate personalized complex event configuration for different types of clients, visual configuration can be performed through an end-to-end intelligent configuration platform, so as to easily configure the matching rules of complex events, making the strategy more flexible and easy to use. With the help of the end-to-end intelligent configuration platform, a simple implementation method of "input configuration" - "output timing matching results" can be achieved.

Next, the experience layering is explained with reference to FIG. 6 and FIG. 7 .

See Figure 6, which is a flow chart of the terminal intelligent experience layering method provided in an embodiment of the present application.

S601: Start the client, send a request for intelligent experience configuration to the server, and obtain an experience layering model;

S602: Start the perception engine to obtain static data of device hardware, real-time status data of the device, and real-time operation data of the scene;

S603: triggering the decision engine to input the acquired static data of device hardware, real-time status data of the device, and real-time operation data of the scenario into the experience layering model to predict the operation status of the device;

S604: Run the contact engine to trigger the corresponding experience layering solution according to the prediction result of the device operation status.

In actual applications, attempts are made to identify the operating status of user devices and to strategically optimize devices in low operating states in order to achieve normal operating status and user experience. Taking mini programs as an example, limited by the mini program container and framework, as well as the complicated scene logic, it is difficult to achieve consistent experience for each model in terms of performance and experience. Based on the long tail theory, the focus of the solution is placed on low-performance devices, with the assistance of a small amount of resources to solve the problem of poor experience. In the existing method, the judgment of low-performance devices is based on the following key factors: hardware specifications, price, performance and functions, etc. These judgment criteria are relatively static and cannot be distinguished more finely according to the real-time operating status of the user's device. For example, after a high-performance mobile phone opens several large-scale game apps, it opens the target mini program, and when the page is very stuck, the traditional judgment method still considers it to be a high-performance device at this moment. The embodiment of the present application will make real-time judgments in combination with the performance of page rendering. In the experience stratification scheme of the embodiment of the present application, the operating status of the user device is evaluated from the perspective of actual user use. Different from the previous rigid evaluation of devices based on factory hardware information, the intelligent body stratification is based on the multi-dimensional time-consuming data of the actual operation of the application, and uses the end-to-end intelligent experience stratification model to perform comprehensive real-time weighted scoring of user device data, and take corresponding measures based on the score.

For the experience stratification solution, it is necessary to first collect and process data on the server side, and train the experience stratification model. Then, the experience stratification model is sent to the front end. During the client operation, based on the acquired target data, the front end uses the experience stratification model to make stratification predictions in real time, and performs scene degradation and other measures based on the operating status of the user's device. The following introduces the solution from two perspectives: data collection and model structure determination.

1. Data Collection and Processing

Data collection and processing are key links in model training. Data processing can be divided into the following processes: feature dimension division, data collection and reporting, data reading, data cleaning, data division, and data preprocessing. These steps provide standard input to the model.

(1) Data dimension division

Different from the existing low-performance device determination scheme, the data in this scheme includes not only static data such as mobile phone model and pixel ratio, but also real-time dynamic data of mobile phone operation, such as current available memory, number of alarms, time taken to load each page, memory changes at each stage, etc. The operation of mobile phones is more real-time, and this scheme can capture the moment of poor operation status in time, and use downgrade measures to prevent the operation status from deteriorating or even improving.

Taking the experience layering model as an example, the collected data can be divided into static data and dynamic data. Static data includes the hardware conditions of the user's mobile phone, such as mobile phone model, configuration, pixel ratio, etc., and dynamic data is divided into device-related data and scenario-related data.

In the embodiment of the present application, the collected data can be divided into the following types: static data of device hardware, real-time status data of device, and real-time operation data of scenes, where static data of device hardware includes parameters such as mobile phone model and memory capacity; real-time status data of device refers to dynamically changing device data, such as parameters such as power or network status; real-time operation data of scenes refers to data related to scene operation, including but not limited to homepage scenes, store scenes, etc., such as parameters such as the time it takes to load the homepage. The embodiment of the present application does not limit the specific parameters of each type of data, such as parameters a, b, c, etc.

(2) Data collection and reporting

The data collection method varies according to the feature dimension. Static data is only read once for caching to reduce time consumption, and dynamic data is extracted at the collection node to ensure accuracy.

Taking the mini program as an example, data reporting methods support different dimensions:

Ordinary tracking data provides daily dimension reports, which are suitable for T+1 offline analysis scenarios;

Highway can transmit data back in minutes, which is suitable for online real-time analysis scenarios;

Answer is relatively flexible and can customize data formats, which is suitable for a wider range of scenarios;

mtop entrainment has high flexibility and real-time performance, but is limited by the interface QPS and trigger timing, and is suitable for end-cloud collaboration scenarios;

(3) Data preprocessing

The quality of data directly determines the prediction effect and generalization ability of the model, and determines the ceiling of the model. Therefore, data processing involves several dimensions: accuracy, completeness, consistency, effectiveness, credibility, and interpretability. Although the data for the experience layering model has undergone basic data cleaning, it still contains a lot of noise.

Therefore, preprocessing is required before the model is built. Standard, clean, and smooth data can be obtained by filling missing values, smoothing noise data, smoothing or deleting outliers, and resolving inconsistent data. In addition, continuous data can be discretized by binning, and continuous data can be divided into corresponding buckets according to certain rules. The binning rules include unsupervised binning (equal frequency binning, equal interval binning), supervised binning (chi-square binning, decision tree binning), etc. Equal frequency and equal interval are commonly used. Combined with the characteristics of experience stratified data, equal frequency binning can be adopted.

2. Model structure and training

In the embodiment of the present application, a convolutional neural network (CNN) can be used to construct an experience layering model. The training process of CNN includes forward propagation (the network calculates the output based on the input data and the current weight), calculating the loss (based on the output and the true label), back propagation (calculating the gradient of each layer based on the loss), and weight update (adjusting the network weight using the gradient descent algorithm or other optimization algorithms).

After further research and exploration, it was found that the first problem to be solved in the experience layering model is the classification problem. Classic algorithms such as GBDT, CNN, and DNN are candidate models. We have tried convolutional neural networks (CNN) and deep neural networks (DNN) successively. They each have their own advantages and disadvantages. CNN is mainly designed to process data with spatial continuity (such as images). By introducing convolution operations and pooling operations, it can effectively process spatial data such as images, but it may not be as efficient as DNN on other types of data. DNN is simpler and more general in structure, but it may face huge parameter space and overfitting problems, but it can be optimized to avoid such problems. Based on the collection of static data of mobile phones and dynamic data such as memory, the comprehensive performance of DNN is better than CNN. Combining the advantages of the two algorithms, in the optimization scheme, CNN can be used for feature extraction, and then the extracted features can be input into DNN for the final model training.

In one implementation, the experience hierarchical model is trained by the "white box + black box model" method. As described above, the weight of the white box model is already specific when it is running online, and there is no need to load the inference model for inference. The white box model is mainly used to screen low-performance devices in the cold start phase when the user opens the client (such as the mini program) to avoid the model running bringing additional burden to the device. This strategy ensures that further operation status diagnosis and experience optimization can be performed through more advanced models in the subsequent process. Black box model: This is a deep network architecture built by methods such as multi-layer nonlinear transformation and abstract global feature learning. It simulates the hierarchical processing of information by the human brain nervous system and is used to solve classification, regression and other problems. Specifically, the experience hierarchical model can be obtained by the "white box model training GBDT structure + black box model training DNN structure" method. The advantage of this method is that first, by using a white box model (such as the GBDT algorithm), the user experience is hierarchically managed according to the operating status of the device. By analyzing the performance parameters of the device (such as memory usage, processor performance, etc.), the model can evaluate the performance level of the device in real time and adjust the performance of the application accordingly, such as reducing the image quality or simplifying the interaction effect to adapt to devices of different performance levels. The decision-making process of the white-box model is transparent, allowing developers to clearly understand the decision logic and basis of the model, which is convenient for tuning and maintenance. Then, by using black-box models (such as DNN), which usually involve complex network structures and a large number of parameters, their internal decision logic is opaque to users. Such models have strong predictive capabilities and are designed to improve the smoothness and personalization of user experience. Therefore, by using the "white box + black box" method to train different network structures to obtain the experience layering model, the characteristics of each model can be brought into play, low-performance devices can be initially screened through the white-box model, and accurate predictions can be made through the black-box model to achieve real-time hierarchical processing solutions.

Therefore, in one implementation, the first experience score model is obtained by training the first network structure (such as GBDT or CNN), and the second experience hierarchical model is obtained by training the second network structure (such as DNN). In the decision-making process, the first experience hierarchical model is used for white box pre-screening and the second experience hierarchical model is used for black box fine screening, so as to predict the operation status of the device. For example, through white box pre-screening, 15% of the devices are initially determined to be in poor operation status and can be downgraded. Then, through black box fine screening, the remaining 85% of the devices are finely determined, and 30% can still be downgraded. This "white box + black box" combined experience stratification method can comprehensively perform device stratification. Experiments have shown that it can significantly save 50% of the devices and reduce the average number of memory alarms by 10%, greatly improving the smoothness of client operation and improving user experience.

In the end-to-end intelligent experience stratification solution, another technical key is to publish and deploy the experience stratification model trained on the server side, so that the client can obtain the experience stratification model during operation and use the model for real-time prediction. Compared with the server or cloud obtaining data from the front end and then performing model prediction, prediction on the front end can reduce the steps of data reporting, thereby speeding up the prediction and reducing latency. Moreover, this model prediction method on the front end is not limited to the network status. Even when the network is stuck or fails, the model prediction can be completed by relying on the front-end data processing capabilities. Therefore, running the experience stratification model on the client side is a technical point.

Referring to Figure 7, a schematic diagram of the logic of implementing the terminal intelligent experience layering provided by the embodiment of the present application is shown. The whole process includes four main steps: model preparation, environment preparation, parameter preparation, and reasoning scoring.

The model preparation stage is to obtain the experience layering model through offline training, such as data collection and processing on the server side and determination of the model structure, so as to use the processed data to train the model structure and obtain the experience layering model.

The environment preparation phase refers to the environment preparation for running the model on the client side, which mainly includes starting the engine. Taking the mini program as an example, first, after the mini program is started, the model is downloaded according to the terminal intelligent configuration information (for example, through the configuration capability of the terminal intelligent configuration platform, the model and related control parameters are configured on the platform, and then the model is saved in a DB, and the model is read from the DB through a configuration gateway service; after the client is started, the client automatically loads the model from the configuration gateway service, please refer to Figure 4 and the description), and then, Tensor is prepared; then, by accessing graphics libraries (such as webGL), programming solutions (such as WASM, a technical solution that uses non-Java programming languages to write code and can run on browsers), image processing (webGPU) and other functions, dynamic switching between the JS engine and the NA engine is achieved.

In the parameter preparation phase, the main work is to prepare model inference parameters, which may include calculation graph parsing, operator loading, weight loading, and loading ready steps.

In the inference operation phase, the inference parameters are mainly input into the model to obtain the prediction results, which may include steps such as Tensor input, input preprocessing, operator operation, and obtaining results.

Therefore, through the above four steps, the experience layering model can be used for real-time reasoning and scoring, thereby determining the real-time operating status of the user's device (such as a mobile phone).

It should be noted that, taking the mini program as an example, since the model file format that can be run in the mini program scenario is limited to .xnn, it is necessary to optimize the model structure and compile heterogeneous models in other formats, which is essentially the conversion of operator operations. Limited by the operating environment and framework of the mini program, the complexity of the model should be simple. The complex model structure is that the model file is large and takes a long time to load, and the calculation is complex and consumes machine resources. Therefore, when designing the model, it is designed in accordance with the principle of simple calculation to avoid the use of complex operators and complex processes.

Therefore, after obtaining the experience layering model, the above method may also include: environment preparation stage: preparing tensors based on the downloaded experience layering model, corresponding to the logic of the model data preprocessing stage in JS, loading the XNN parser to implement dynamic JS engine and NA engine; parameter preparation stage: preparing model parameters based on the model format supported by the client, including model structure optimization and heterogeneous model compilation.

Model structure optimization refers to the process of improving efficiency and reducing resource consumption by improving the architecture of the model. This usually involves the following aspects: Pruning: Removing unimportant weights or neurons to reduce model size. Quantization: Reducing the precision of model parameters to reduce model size and accelerate reasoning, such as converting from 32-bit floating point numbers to 8-bit integers. Simplification: Simplify the model and remove overly complex parts to improve operating efficiency. Knowledge Distillation: Transferring the knowledge of a large model to a smaller model. AutoML for model compression: Use automated algorithms to find the optimal model compression method.

Heterogeneous model compilation: Closely related to model structure optimization, it refers to the process of compiling deep learning models for different hardware platforms (such as GPU, CPU, TPU or FPGAs, etc.). With the diversification of hardware devices, a model that runs efficiently on a specific hardware may not perform well on another type of hardware. Heterogeneous compilation achieves model portability and performance optimization by converting the model to adapt it to the characteristics of different hardware.

In summary, model structure optimization focuses on reducing complexity and improving efficiency within the model, while heterogeneous model compilation focuses on model portability and performance optimization across different hardware platforms. These two aspects are critical in deploying deep learning models into actual production environments.

In one implementation, the experience layering model is pre-trained on the server side and published to the content distribution network, wherein the decision engine accesses the content distribution network and loads the experience layering model locally according to the configuration information of the experience layering model. For example, the model and related control parameters are configured on the platform through the configuration capability of the end intelligent configuration platform, and then the model is saved in a DB, and the model is read from the DB through a configuration gateway service; after the client is started, the client automatically loads the model from the configuration gateway service, as shown in Figure 4 and the description.

For example, the prediction results of the experience stratification model are divided into three levels: good running status, general running status, and poor running status. For the latter two levels, path simplification, rich interaction degradation, and function degradation can be adopted. Path simplification can be to shut down non-critical functions or reduce the priority of non-critical functions. For example, if non-critical functions have no effect on users placing orders or completing order fulfillment, these functions can be directly shut down, or if non-critical functions have little effect on users placing orders or completing order fulfillment, the priority of these functions can be reduced. Rich interaction degradation can include reducing animation frame rate, reducing image resolution, simplifying interaction effects, etc. Reduce animation frame rate: The higher the animation frame rate, the higher the machine performance requirements. Therefore, the animation frame rate can be adjusted in real time according to the performance of the user's machine to achieve a smooth running effect; reduce image resolution: adjust the image resolution in real time according to the performance of the user's machine to reduce the memory consumption of image rendering; simplify interaction effects: Some interaction effects have high requirements for machine performance. These interaction effects (such as lottie animation effects) can be simplified or turned off in real time according to the performance of the user's machine. Function degradation refers to the degradation of non-blocking user order placement or order fulfillment functions in the client.

The experience stratification scheme provided in the embodiment of the present application is a cross-end, instant, dynamic, intelligent, and full-link low-performance device identification strategy, which not only focuses on the static data of the device, but also starts from the various dimensional data during the operation of the device, and provides more accurate data basis for the determination of low-performance devices through real-time perception calculation. Compared with the traditional way of evaluating low-performance devices, the traditional method is more rigid and not flexible enough, while the embodiment of the present application focuses on real-time data analysis and algorithm decision-making, starting from the real user experience, to provide a more intelligent, comprehensive and accurate evaluation scheme, which breaks through the limitations of traditional schemes and is more personalized, intelligent and immediate.

Finally, the experience enhancement solution is introduced in conjunction with FIG8 .

Referring to FIG8 , a flow chart of a method for enhancing terminal intelligent experience provided in an embodiment of the present application is shown.

S801: Start the client, send a request for intelligent experience configuration to the server, and obtain a full-link traffic flow prediction model;

S802: Start the perception engine to obtain user behavior data, user basic data, scenario data and performance data;

S803: triggering the decision engine to input the acquired user behavior data, user basic data, scenario data, and performance data into the full-link traffic flow prediction model to predict the user's future behavior trajectory;

S804: Run the contact engine to trigger a corresponding experience enhancement solution based on the prediction result of the user's future behavior trajectory.

Experience enhancement refers to the use of a full-link traffic flow prediction model based on real-time user interactive behavior information to achieve real-time prediction of users' future trajectories, preload sub-packages and pages that users may visit, and enhance the user experience.

For the experience enhancement solution, it is necessary to first collect and process data on the server side, and train the full-link traffic flow prediction model. Then, the full-link traffic flow prediction model is sent to the front end. During the client operation, based on the acquired target data, the full-link traffic flow prediction model is used to predict future behaviors in real time on the front end, and based on the future behavior prediction results, sub-packaging and preloading are performed.

Regarding the two aspects of data collection and model structure determination, the principles and processing procedures of the experience layering scheme are similar, and reference can be made to the relevant description of the experience layering scheme. The difference lies in the division of data feature dimensions and the selection of model structure. Since experience enhancement mainly considers predicting future user behavior, the data feature dimensions mainly include user behavior data, user basic data, scenario data, and performance data. For various data acquisition methods, reference can be made to the previous description. For the network structure of the full-link dynamic line prediction model, through research and practice, a network structure with strong computing power can be selected, such as a deep learning network, such as LSTM, etc. This model analyzes the user's historical behavior data and tries to predict the user's next possible operation or preference, so as to predict the user's behavior path. Since deep learning models usually involve complex network structures and a large number of parameters, their internal decision logic is opaque to users, so a black box model can be used. Therefore, in one implementation method, the full-link dynamic line prediction model corresponding to the experience enhancement decision is set to a black box model mode, and in the client hot start phase, the full-link dynamic line prediction model in the black box model mode is run.

Regarding publishing and deploying the full-link traffic flow prediction model trained on the server, as well as downloading it on the client and performing real-time prediction locally, please refer to the description of the experience tiering solution (Figure 7). The principles and implementation process are similar and will not be repeated here.

In addition, in one implementation, the full-link traffic prediction model is pre-trained on the server and published to the content distribution network, wherein the decision engine accesses the content distribution network and loads the experience layering model locally based on the configuration information of the experience layering model. For example, through the configuration capability of the end-to-end intelligent configuration platform, the model and related control parameters are configured on the platform, and then the model is saved in a DB, and the model is read from the DB through a configuration gateway service; after the client is started, the client automatically loads the model from the configuration gateway service, as shown in Figure 4 and the description.

For example, the user's future behavior trajectory can be obtained through the full-link traffic flow prediction model, and corresponding experience enhancement measures can be taken based on the prediction results. For example, the experience enhancement measures include: obtaining the user's future behavior trajectory based on the experience decision, determining the sub-package resources or page resources that the user may access, and preloading the sub-package resources or page resources.

In a specific scenario, intelligent preloading can be performed for channel sub-packaging. In the client interface, the next hop page of the homepage is usually sub-packaging. Hard full preloading not only wastes traffic, but also occupies memory and even causes jamming. By using the end-to-end intelligent experience enhancement solution to predict the next hop address and preload sub-packaging according to the principle of proximity, the contradiction between memory and experience can be better balanced.

Corresponding to the above-mentioned method for implementing the end intelligent experience applied on the client, the embodiment of the present application also provides a method for implementing the end intelligent experience applied on the server. See Figure 9, which is a flow chart of the method for implementing the end intelligent experience applied on the server.

The end intelligent experience implementation method is used to configure and touch the end intelligent experience front-end code, wherein the front-end code is applicable to multiple types of clients and performs perception, decision-making and processing of user experience at the front end. The method is applied to the server, including:

S901: receiving a terminal intelligent configuration request sent by a client, matching the terminal intelligent configuration and sending it to the client, so that the client executes the terminal intelligent experience decision and processing;

The client executes experience decision and disposal, including: running the end intelligent suite in the front-end code, including: starting the perception engine in the end intelligent suite to obtain target data in real time; triggering the decision engine in the end intelligent suite to make experience decisions for the target data based on the end intelligent experience configuration; and running the contact engine in the end intelligent suite to trigger the corresponding disposal plan according to the experience decision result.

S902: In response to the triggering of the callback function, receiving the processing solution execution result fed back by the client.

In one implementation, the method further includes: configuring the end intelligent experience through the end intelligent configuration platform of the server, including configuring complex event configuration information, experience layering model and/or full-link traffic flow prediction model.

In one implementation, the method further includes: receiving an experience handling solution and user feedback data from the client as a new input-end intelligent configuration platform to optimize complex event configuration information, an experience layering model and/or a full-link traffic flow prediction model.

In one implementation, the method further includes: sending network resources corresponding to the handling solution to the client. For example, in an experience enhancement scenario, sending sub-packet resources to the client in advance.

For the implementation method of the end-to-end intelligent experience executed on the server side, please refer to the previous principles and descriptions, which will not be repeated here.

Referring to FIG. 10 , there is shown a schematic diagram of the structure of a device for implementing a terminal intelligent experience provided in an embodiment of the present application.

The device is used to perceive, decide and handle user experience at the front end for front end codes applicable to multiple types of clients. The device is applied to one client of any type of client among the multiple types of clients, and includes:

The configuration request unit 1001 is used to start the client, send a terminal intelligent experience configuration request to the server, obtain the terminal intelligent experience configuration and store it locally;

The end intelligent execution unit 1002 is used to run the end intelligent suite in the front-end code, including: starting the perception engine in the end intelligent suite to obtain the target data in real time; triggering the decision engine in the end intelligent suite to make experience decisions for the target data based on the end intelligent experience configuration; and running the contact engine in the end intelligent suite to trigger the corresponding disposal plan according to the experience decision result.

In one implementation, the processing scheme includes: an experience monitoring processing scheme, an experience layering processing scheme, and/or an experience enhancement processing scheme;

The experience monitoring and handling scheme includes: when a complex event is determined to occur according to the experience decision, real-time message push and/or page jump operation is performed in the interface in a UI interactive manner;

The experience layering solution includes: when the device operation state is determined to be a low operation state according to the experience decision, performing interface loading optimization and/or page content dynamic adjustment solution;

The experience enhancement solution includes: obtaining a user's future behavior trajectory based on the experience decision, determining sub-package resources or page resources that the user may access, and preloading the sub-package resources or page resources.

In one implementation, the method further includes:

The handling collaborative control unit 1003 is used to set different handling schemes corresponding to different priorities, and determine the execution logic relationship between different handling schemes according to the priorities; wherein, the experience stratification handling scheme is set to correspond to the first priority; according to the processing results of the experience stratification handling scheme, determine whether to execute the experience monitoring handling scheme and/or the experience enhancement handling scheme.

In one implementation, the method further includes:

The collaborative control unit 1003 performs experience monitoring, experience stratification or experience enhancement experience decisions through a white box model or a black box model, makes preliminary decisions based on the white box model, and determines whether to use the black box model to make further refined decisions based on the preliminary decision results of the white box model.

In one implementation, the treatment coordination control unit 1003 is specifically used to:

Setting the experience layering model corresponding to the experience layering decision to a white box model mode, running the experience layering model in the white box model mode during the cold start phase of the client, and determining whether to continue to execute the terminal intelligence suite or shut down the terminal intelligence suite according to the prediction result of the experience layering model; and

The full-link traffic flow prediction model corresponding to the experience enhancement decision is set to a black box model mode, and during the client hot start phase, the full-link traffic flow prediction model of the black box model mode is run.

In one implementation, the terminal intelligent experience configuration includes complex event configuration information, and the complex event configuration information includes event configuration information and timing configuration information of at least one target complex event;

The end intelligent execution unit 1002 is specifically used for: starting the perception engine, capturing the embedded data based on the unified event stream processing tool in the front-end code, and determining whether the target complex event corresponding to the event configuration information is triggered according to the embedded data; in response to the triggering of the target complex event, running the decision engine to match the timing for the embedded data according to the timing configuration information to determine whether the timing of the target complex event is met; and, after the timing is successfully matched, running the contact engine to trigger the experience monitoring and disposal plan corresponding to the target complex event.

In one implementation, the unified event stream processing tool is specifically used to:

Perform client type analysis and format analysis on the buried data, and standardize the buried data so that the data of different types of clients have a unified format and structure; wherein the standardized processing of the buried data includes: parsing the buried data's buried information, event ID, and parameter information; packaging the buried information, event ID, and parameter information in a unified format, and converting them into event information in a standard format.

In one implementation, the decision engine is a complex event processing component based on a self-developed complex event matching rule library and declared as a static function, wherein the self-developed complex event matching rule library is developed to be compatible with multiple types of clients.

In one implementation, the terminal intelligent experience configuration includes an experience layering model;

The terminal intelligent execution unit 1002 is specifically used for: starting the perception engine, acquiring the static data of the device hardware, the real-time status data of the device, and the real-time operation data of the scene; inputting the acquired static data of the device hardware, the real-time status data of the device, and the real-time operation data of the scene into the experience layering model to predict the operation status of the device; and running the contact engine to trigger the corresponding experience layering disposal plan according to the prediction result of the operation status of the device.

In one implementation, the experience layering model is obtained by extracting features from data samples based on a convolutional neural network and inputting the extracted features into a deep neural network model for training.

In one implementation, the experience layering model is obtained by training a first network structure to obtain a first experience model and training a second network structure to obtain a second experience layering model. In the decision-making process of the decision engine, white box pre-screening is performed through the first experience layering model and black box fine screening is performed through the second experience layering model to predict the operating status of the device.

In one implementation, the terminal intelligent experience configuration includes a full-link dynamic line prediction model;

The terminal intelligent execution unit 1002 is specifically used to: start the perception engine to obtain user behavior data, user basic data, scenario data and performance data; input the obtained user behavior data, user basic data, scenario data and performance data into the full-link traffic flow prediction model to predict the user's future behavior trajectory; and run the contact engine to trigger the corresponding experience enhancement disposal plan according to the prediction results of the user's future behavior trajectory.

An embodiment of the present application further provides a storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

Optionally, in this embodiment, the storage medium may be configured to store a computer program for performing the following steps:

Start the client, send a request for the client intelligent experience configuration to the server, obtain the client intelligent experience configuration and store it locally;

Running the end intelligence suite in the front-end code includes: starting the perception engine in the end intelligence suite to obtain target data in real time; triggering the decision engine in the end intelligence suite to make experience decisions for the target data based on the end intelligence experience configuration; and running the contact engine in the end intelligence suite to trigger corresponding disposal plans based on the experience decision results.

Optionally, in this embodiment, the above-mentioned storage medium may include but is not limited to: a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and other media that can store computer programs.

An embodiment of the present application further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

Optionally, the electronic device may further include a transmission device and an input/output device, wherein the transmission device is connected to the processor, and the input/output device is connected to the processor.

Optionally, in this embodiment, the processor may be configured to perform the following steps through a computer program:

Start the client, send a request for the client intelligent experience configuration to the server, obtain the client intelligent experience configuration and store it locally;

Running the end intelligence suite in the front-end code includes: starting the perception engine in the end intelligence suite to obtain target data in real time; triggering the decision engine in the end intelligence suite to make experience decisions for the target data based on the end intelligence experience configuration; and running the contact engine in the end intelligence suite to trigger corresponding disposal plans based on the experience decision results.

Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments and optional implementation modes, and this embodiment will not be described in detail here.

The serial numbers of the embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc., which can store program code.

The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims (15)

1. A method for realizing intelligent user experience on a client side, characterized in that it is used to perceive, decide and handle user experience on the front end for front end codes applicable to multiple types of clients, and the method is applied to one client of any type of client among the multiple types of clients, comprising: Start the client, send a terminal intelligent experience configuration request to the server, obtain the terminal intelligent experience configuration and store it locally, wherein the terminal intelligent experience configuration includes event configuration information and/or model configuration information; Running the end intelligence suite in the front-end code includes: starting the perception engine in the end intelligence suite to obtain target data in real time; triggering the decision engine in the end intelligence suite to make experience decisions for the target data according to the end intelligence experience configuration; and running the touchpoint engine in the end intelligence suite to trigger corresponding disposal plans according to the experience decision results; The processing scheme includes an experience monitoring processing scheme, an experience layering processing scheme and/or an experience enhancement processing scheme; The experience monitoring and handling scheme includes: when a complex event is determined to occur according to the experience decision, real-time message push and/or page jump operation is performed in the interface in a UI interactive manner; The experience layering solution includes: when the device operation state is determined to be a low operation state according to the experience decision, performing interface loading optimization and/or page content dynamic adjustment solution; The experience enhancement solution includes: obtaining a user's future behavior trajectory based on the experience decision, determining sub-package resources or page resources that the user may access, and preloading the sub-package resources or page resources. 2. The method according to claim 1, further comprising: Set different handling plans to correspond to different priorities, and determine the execution logic relationship between different handling plans according to the priorities; among them, set the experience stratification handling plan to correspond to the first priority; according to the processing results of the experience stratification handling plan, determine whether to execute the experience monitoring handling plan and/or the experience enhancement handling plan. 3. The method according to claim 1, further comprising: Use white box models or black box models to make experience decisions for experience monitoring, experience layering, or experience enhancement, and make preliminary decisions based on the white box model, and determine whether to use the black box model to make further refined decisions based on the preliminary decision results of the white box model. 4. The method according to claim 3, further comprising: Setting the experience layering model corresponding to the experience layering decision to a white box model mode, running the experience layering model in the white box model mode during the cold start phase of the client, and determining whether to continue to execute the terminal intelligence suite or shut down the terminal intelligence suite according to the prediction result of the experience layering model; The full-link traffic flow prediction model corresponding to the experience enhancement decision is set to a black box model mode, and during the client hot start phase, the full-link traffic flow prediction model of the black box model mode is run. 5. The method according to any one of claims 1 to 4, characterized in that the terminal intelligent experience configuration includes complex event configuration information, and the complex event configuration information includes event configuration information and timing configuration information of at least one target complex event; The perception engine in the startup end intelligent suite acquires target data in real time, including: starting the perception engine, capturing embedded point data based on the unified event stream processing tool in the front-end code, and determining whether the target complex event corresponding to the event configuration information is triggered according to the embedded point data; The decision engine in the trigger end intelligent suite makes an experience decision for the target data according to the end intelligent experience configuration, including: in response to the triggering of the target complex event, running the decision engine to match the timing of the buried point data according to the timing configuration information to determine whether the timing of the target complex event is met; The contact engine in the operating end intelligent suite triggers a corresponding disposal plan according to the experience decision result, including: after the timing is successfully matched, running the contact engine to trigger the experience monitoring disposal plan corresponding to the target complex event. 6. The method according to claim 5, characterized in that after the unified event stream processing tool captures the tracking point data, it also includes: Perform client type and format analysis on the tracking data, and standardize the tracking data so that the data of different types of clients have a unified format and structure; Among them, the standardized processing of the buried point data includes: parsing the buried point information, event ID, and parameter information of the buried point data; packaging the buried point information, event ID, and parameter information in a unified format, and converting them into event information in a standard format. 7. The method according to claim 5, characterized in that The decision engine is a complex event processing component based on a self-developed complex event matching rule library and declared as a static function, wherein the self-developed complex event matching rule library is developed using JavaScript to be compatible with multiple types of clients. 8. The method according to any one of claims 1 to 4, characterized in that the terminal intelligent experience configuration includes an experience layering model; The perception engine in the startup end intelligent kit obtains target data in real time, including: starting the perception engine to obtain static data of device hardware, real-time status data of the device, and real-time operation data of the scene; The decision engine in the trigger end intelligent suite makes an experience decision for the target data according to the end intelligent experience configuration, including: triggering the decision engine to input the acquired static data of device hardware, real-time status data of the device, and real-time operation data of the scene into the experience hierarchical model to predict the operation status of the device; The touchpoint engine in the operation-end intelligent suite triggers a corresponding disposal plan according to the experience decision result, including: the operation touchpoint engine triggers a corresponding experience layered disposal plan according to the prediction result of the device operation status. 9. The method according to claim 8 is characterized in that the experience layering model is obtained by extracting features from data samples based on a convolutional neural network and inputting the extracted features into a deep neural network model for training. 10. The method according to claim 8, characterized in that The experience layering model is obtained by training a first network structure to obtain a first experience layering model and training a second network structure to obtain a second experience layering model. In the decision-making process of the decision engine, the first experience layering model is used as a white box model for white box pre-screening and the second experience layering model is used as a black box model for black box fine screening, so as to predict the operating status of the device. 11. The method according to any one of claims 1 to 4, characterized in that the terminal intelligent experience configuration includes a full-link dynamic line prediction model; The perception engine in the startup terminal intelligent suite acquires target data in real time, including: starting the perception engine to acquire user behavior data, user basic data, scenario data and performance data; The decision engine in the trigger end intelligent suite makes experience decisions for target data according to the end intelligent experience configuration, including: triggering the decision engine to input the acquired user behavior data, user basic data, scenario data and performance data into the full-link dynamic line prediction model to predict the user's future behavior trajectory; The touch engine in the operating end intelligent suite triggers a corresponding disposal plan according to the experience decision result, including: running the touch engine to trigger a corresponding experience enhancement disposal plan according to the prediction result of the user's future behavior trajectory. 12. A method for implementing an end intelligent experience, characterized in that it is used to configure and contact the end intelligent experience front-end code, wherein the front-end code is applicable to multiple types of clients and performs perception, decision-making and processing of user experience at the front end, and the method is applied to the server, including: Receive an end-intelligent configuration request sent by a client, match an end-intelligent configuration and send it to the client, wherein the end-intelligent experience configuration includes event configuration information and/or model configuration information, so that the client performs the following steps: run the end-intelligent suite in the front-end code, wherein the perception engine in the end-intelligent suite is started to obtain target data in real time; trigger the decision engine in the end-intelligent suite, and make an experience decision for the target data according to the end-intelligent experience configuration; and run the contact engine in the end-intelligent suite to trigger a corresponding disposal plan according to the experience decision result, wherein the disposal plan includes an experience monitoring disposal plan, an experience layering disposal plan and/or an experience enhancement disposal plan, wherein the experience monitoring disposal plan includes: when a complex event is determined to occur according to the experience decision, real-time message push and/or page jump operation are performed in the interface in a UI interactive manner; the experience layering disposal plan includes: when the device operation state is determined to be a low operation state according to the experience decision, interface loading optimization and/or page content dynamic adjustment plan are performed; the experience enhancement disposal plan includes: obtaining the user's future behavior trajectory according to the experience decision, determining the sub-package resources or page resources that the user may access, and preloading the sub-package resources or page resources; and, In response to the triggering of the callback function, the execution result of the disposal plan fed back by the client is received. 13. A device for realizing intelligent user experience, characterized in that it is used to perceive, decide and handle user experience at the front end for front end codes applicable to multiple types of clients, and the device is applied to one client of any type of client among the multiple types of clients, comprising: A configuration request unit, used to start the client, send a terminal intelligent experience configuration request to the server, obtain the terminal intelligent experience configuration and store it locally, wherein the terminal intelligent experience configuration includes event configuration information and/or model configuration information; The end intelligent execution unit is used to run the end intelligent suite in the front-end code, including: starting the perception engine in the end intelligent suite to obtain the target data in real time; triggering the decision engine in the end intelligent suite to make an experience decision for the target data according to the end intelligent experience configuration; and running the touchpoint engine in the end intelligent suite to trigger the corresponding disposal plan according to the experience decision result; Among them, the handling scheme includes an experience monitoring handling scheme, an experience layering handling scheme and/or an experience enhancement handling scheme. The experience monitoring handling scheme includes: when a complex event is determined to occur according to the experience decision, real-time message push and/or page jump operation is performed in the interface in a UI interactive manner; the experience layering handling scheme includes: when the device operating state is determined to be a low operating state according to the experience decision, interface loading optimization and/or page content dynamic adjustment scheme is performed; the experience enhancement handling scheme includes: obtaining the user's future behavior trajectory according to the experience decision, determining the sub-package resources or page resources that the user may access, and preloading the sub-package resources or page resources. 14. A storage medium, characterized in that a computer program is stored in the storage medium, wherein the computer program is configured to execute the method according to any one of claims 1 to 12 when running. 15. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the method according to any one of claims 1 to 12.