CN112070542A - Information conversion rate prediction method, apparatus, device and readable storage medium - Google Patents

Abstract

The invention discloses an information conversion rate prediction method, which comprises the following steps: obtaining expanded user data, reconstructing original training data based on the expanded user data to obtain enhanced training data, performing combined model training to obtain an information conversion rate prediction model and an auxiliary model, optimizing the information conversion rate prediction model based on the auxiliary model, and obtaining a conversion rate prediction result of information content corresponding to the information conversion rate prediction request according to the optimized information conversion rate prediction model when the information conversion rate prediction request is received. The invention also discloses a device, equipment and a readable storage medium. By introducing the auxiliary model based on enhanced training data training in the combined model training process, the prediction effect that the information conversion rate prediction model is compatible with the auxiliary model is achieved, so that the high-quality information conversion rate prediction model is obtained, the online information delivery prediction accuracy is improved, and the success rate of accurate information delivery is improved.

Description

Information conversion rate prediction method, apparatus, device and readable storage medium

technical field

The present invention relates to the artificial intelligence technology field of financial technology (Fintech), and in particular, to an information conversion rate prediction method, device, device and readable storage medium.

Background technique

With the continuous development of financial technology, especially Internet technology finance, more and more technologies are applied in the financial field, but the financial industry also puts forward higher requirements for technology, such as the financial industry also has higher requirements for online information promotion .

The pushed information includes advertisements. Online advertisements have become a common way of advertising, and conversion rate estimation is an important module for accurate online advertisements. Based on the machine learning algorithm, the estimation module uses the user's basic attribute information, behavior data, and analyzes the advertisement content data to find the most suitable advertisement for the user.

The conversion rate estimation of existing online advertisements generally only uses the user information collected by the system in the early stage, such as the user's basic attribute information, user interest information, etc., as well as the user's conversion behavior data to train the conversion rate estimation model. Due to the insufficient amount of data for conversion rate estimation, the accuracy of the conversion rate prediction of advertisements is not high, which affects the accurate delivery of advertisements.

SUMMARY OF THE INVENTION

The main purpose of the present invention is to provide an information conversion rate prediction method, device, device and readable storage medium, which aims to solve the technical problem that the conversion rate prediction accuracy of existing online information is not high, which affects the accurate delivery of information.

In order to achieve the above object, the present invention provides a method for predicting information conversion rate, the method for predicting information conversion rate includes:

Obtaining extended user data corresponding to the user equipment number of the original training data, and reconstructing the original training data based on the extended user data to obtain enhanced training data;

Use the original training data and the enhanced training data to perform joint model training to obtain an information conversion rate prediction model and an auxiliary model, and optimize the information conversion rate prediction model based on the auxiliary model;

When an information conversion rate prediction request is received, a conversion rate prediction result of the information content corresponding to the information conversion rate prediction request is obtained according to the optimized information conversion rate prediction model.

Further, the original training data includes the original conversion rate corresponding to the original information feature data and the user equipment number;

The step of obtaining the expanded user data corresponding to the user equipment number of the original training data, and reconstructing the original training data based on the expanded user data, the steps of obtaining the enhanced training data include:

In the preset user assistance information database, look up the extended user data according to the user equipment number, and in the preset user portrait database, look up the user portrait data according to the user equipment number;

Based on the user equipment number, the extended user data, the user portrait data, the original information feature data, and the original conversion rate are associated and saved as the enhanced training data.

Further, using the original training data and the enhanced training data to perform joint model training to obtain an information conversion rate prediction model and an auxiliary model, and optimize the information conversion rate prediction model based on the auxiliary model. Steps include:

Input the original training data into the initial prediction model for model training, while inputting the enhanced training data into the auxiliary model for model training, based on the first prediction value of the initial prediction model and the second prediction value of the auxiliary model , construct the first total loss function corresponding to the initial prediction model and the second total loss function of the auxiliary model;

Perform model iterative training based on the first total loss function and the second total loss function, and use the second predicted value of the auxiliary model to optimize the initial prediction model;

When the initial prediction model converges, the current initial prediction model is determined as the optimized information conversion rate prediction model.

Further, the step of constructing the first total loss function corresponding to the initial prediction model based on the first predicted value of the initial prediction model and the second predicted value of the auxiliary model includes:

determining a first loss function based on the first predicted value of the initial prediction model and the original conversion rate of the original training data;

determining a third loss function based on the first predicted value of the initial prediction model and the second predicted value of the auxiliary model;

Weighted summation is performed on the first loss function and the third loss function to obtain the first total loss function.

Further, when the initial prediction model converges, the step of determining the current initial prediction model as the optimized information conversion rate prediction model includes:

determining whether the initial prediction model converges based on the first total loss function and a preset threshold;

When the first total loss function is less than or equal to a preset threshold, it is determined that the initial prediction model is converged, and the optimized prediction model of the information conversion rate is obtained.

Further, the step of constructing a second total loss function corresponding to the auxiliary model based on the first predicted value of the initial prediction model and the second predicted value of the auxiliary model includes:

determining a second loss function based on the second predicted value of the auxiliary model and the original conversion rate of the original training data;

A weighted summation is performed on the third loss function and the second loss function determined based on the first predicted value of the initial prediction model and the second predicted value of the auxiliary model, to obtain the second total loss function.

Further, the information conversion rate prediction request includes the target user equipment number and target information feature data, and when the information conversion rate prediction request is received, the information conversion rate is obtained according to the optimized information conversion rate prediction model. The steps of predicting the conversion rate prediction result of the information content corresponding to the request include:

In the preset user portrait database, search for target user portrait data according to the target user equipment number;

Inputting the target user portrait data and the target information feature data into the optimized information conversion rate prediction model for conversion rate prediction, and using the predicted value output by the information conversion rate prediction model as the predicted conversion rate.

Further, the information conversion rate prediction device includes:

a building module for acquiring extended user data corresponding to the user equipment number of the original training data, and reconstructing the original training data based on the extended user data to obtain enhanced training data;

A training module, configured to use the original training data and the enhanced training data to perform joint model training, obtain an information conversion rate prediction model and an auxiliary model, and optimize the information conversion rate prediction model based on the auxiliary model;

The prediction module is configured to obtain, when an information conversion rate prediction request is received, a conversion rate prediction result of the information content corresponding to the information conversion rate prediction request according to the optimized information conversion rate prediction model.

In order to achieve the above object, the present invention also provides an information conversion rate prediction device, the device includes: a memory, a processor and an information conversion rate prediction program stored on the memory and running on the processor, When the information conversion rate prediction program is executed by the processor, the steps of the above-mentioned information conversion rate prediction method are realized.

In addition, in order to achieve the above object, the present invention also provides a readable storage medium, on which an information conversion rate prediction program is stored, and when the information conversion rate prediction program is executed by a processor, any one of the above-mentioned items is realized. The steps of the information conversion rate prediction method.

The present invention obtains the extended user data corresponding to the user equipment number of the original training data, and reconstructs the original training data based on the extended user data to obtain enhanced training data, and then utilizes the original training data and the enhanced training data. The information conversion rate prediction model and the auxiliary model are obtained, and the information conversion rate prediction model is optimized based on the auxiliary model. The information conversion rate prediction model obtained by the information conversion rate prediction model obtains the conversion rate prediction result of the information content corresponding to the information conversion rate prediction request. By introducing an auxiliary model based on enhanced training data in the joint model training process, the information conversion rate prediction model can be compatible with the prediction effect of the auxiliary model, so as to obtain a high-quality information conversion rate prediction model and improve the accuracy of online information delivery prediction. rate, thereby improving the success rate of accurate information delivery.

Description of drawings

1 is a schematic structural diagram of a device in a hardware operating environment involved in an embodiment of the present invention;

Fig. 2 is the schematic flow chart of the first embodiment of the information conversion rate prediction method of the present invention;

3 is a schematic flowchart of the second embodiment of the information conversion rate prediction method of the present invention;

FIG. 4 is a schematic diagram of functional modules of an embodiment of an information conversion rate prediction apparatus according to the present invention.

The realization, functional characteristics and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments.

Detailed ways

It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.

As shown in FIG. 1 , FIG. 1 is a schematic structural diagram of a device in a hardware operating environment involved in an embodiment of the present invention.

As shown in FIG. 1 , the device may include: a processor 1001 , such as a CPU, a network interface 1004 , a user interface 1003 , a memory 1005 , and a communication bus 1002 . Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. Optionally, the network interface 1004 may include a standard wired interface and a wireless interface (eg, a WI-FI interface). The memory 1005 may be high-speed RAM memory, or may be non-volatile memory, such as disk memory. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001 .

Those skilled in the art can understand that the device structure shown in FIG. 1 does not constitute a limitation on the device, and may include more or less components than the one shown, or combine some components, or arrange different components.

As shown in FIG. 1 , the memory 1005 as a readable storage medium may include an operating system, a network communication module, a user interface module, and an information conversion rate prediction program.

In the device shown in FIG. 1 , the network interface 1004 is mainly used to connect to the background server and perform data communication with the background server; the user interface 1003 is mainly used to connect to the client and perform data communication with the client; and the processor 1001 can be used for The information conversion rate prediction program stored in the memory 1005 is called.

In this embodiment, the device includes: a memory 1005, a processor 1001, and an information conversion rate prediction program stored on the memory 1005 and executable on the processor 1001, wherein the processor 1001 calls the memory 1005 to store the When the information conversion rate prediction program is created, the steps of the information conversion rate prediction method provided by each embodiment of the present application are executed.

The present invention also provides a method for predicting an information conversion rate. Referring to FIG. 2 , FIG. 2 is a schematic flowchart of a first embodiment of the method for predicting an information conversion rate according to the present invention.

This embodiment of the present invention provides an embodiment of a method for predicting an information conversion rate. It should be noted that although a logical sequence is shown in the flowchart, in some cases, the sequence shown here may be performed in a different order. or the described steps.

The information conversion rate prediction method of the present invention is applicable to various online information, such as advertisements, electronic journals, news, etc. For the convenience of description, advertisements are used as an example for description in each embodiment of the present invention.

In this embodiment, the information conversion rate prediction method includes:

Step S10, acquiring extended user data corresponding to the user equipment number of the original training data, and reconstructing the original training data based on the extended user data to obtain enhanced training data;

In this embodiment, after clicking on the advertisement, the user may purchase the product in the advertisement, or download the application in the advertisement, etc. This behavior is called conversion, and the probability that the user will generate conversion behavior after the advertisement is predicted is called as Conversion rate (post-click conversion rate, cvr) estimates. Conversion rate estimation is a very important task in the field of online advertising. It has important applications in achieving accurate online advertising and evaluating the value of advertising. An accurate and easy-to-use conversion rate estimation model is very important for advertising platforms and advertisers. A win-win situation is of great significance, in which the advertiser refers to the party that invests in advertising, and the advertising platform refers to the party that uses its own platform technology to help the advertiser screen out the target users of the advertisement.

As mentioned above, the conversion rate estimation of existing online advertisements generally only uses the user information collected by the system in the early stage, such as the user's basic attribute information, user interest information, etc., as well as the user's conversion behavior data to train the conversion rate estimation model. Due to the insufficient amount of data for conversion rate estimation, the accuracy of the conversion rate prediction of advertisements is not high, which affects the accurate delivery of advertisements. Therefore, the advertising conversion rate prediction method proposed by the present invention can obtain a high-quality information conversion rate prediction model by constructing enhanced training data based on the expanded user data for joint model training, thereby improving the prediction accuracy rate of online advertising, thereby improving the accuracy of advertising. Delivery success rate. Among them, extended user data refers to some background information of users. For example, after users leave their company information, we can obtain relevant information of a large number of enterprises based on the data, such as the industry to which the company belongs, the place of registration, the time of registration, and the scale of the company.

Further, the original training data includes a user equipment number, and the corresponding extended user data is obtained according to the user equipment number of the original training data, and the original training data is reconstructed by using the extended user data, thereby obtaining enhanced training data.

Specifically, step S10 includes:

Step S11, in a preset user assistance information database, search for the extended user data according to the user equipment number, and in a preset user portrait database, search for user portrait data according to the user equipment number; User portrait data;

Step S13: Based on the user equipment number, the extended user data, the user portrait data, the original information feature data, and the original conversion rate are associated and saved as the enhanced training data.

In this embodiment, the original training data includes the user equipment number, the original advertisement feature data, and the original conversion rate corresponding to the user equipment number. Among them, the user equipment number is the device code of the mobile phone or other intelligent terminal, the English abbreviation is: IMEI, that is, the International Mobile Equipment Identity Code, which is composed of 15 digits. In layman's terms, the device number is the ID card of the smart terminal, which is assigned at the factory and is unique among mobile devices in the world. Usually, the device code of a smart terminal is owned by a single user, so in the present invention, the user device number is used as an identifier, that is, a user device number represents a user, and which user device numbers can be determined according to the predicted conversion rate. is the object of advertisement; the original advertisement feature data is the attribute characteristics of the advertisement, including but not limited to the characteristics of the advertisement material, the contextual keywords of the advertisement, etc.; the original conversion rate corresponding to the user equipment number indicates whether the user purchases the product in the advertisement, If the product is purchased, it means that the conversion has occurred. If the product is not purchased, it means that the conversion has not occurred. You can set the original conversion rate to 1, which means that the conversion has occurred, and when the original conversion rate is 0, it means that the conversion has not occurred.

Specifically, the preset user portrait database is used to store user portrait information, including but not limited to the user's basic attribute information, interest characteristics and other data, such as the user's gender, age, location, shopping preferences, reading interests, and the like. In the user portrait database, the user portrait information is stored in a one-to-one correspondence with the user device number, so the user device number can be used to search for the user portrait information corresponding to the user device number in the preset user portrait database. In the same way, the user auxiliary information database is used to store user background information. For example, after users leave their company information, we can obtain a large number of relevant information of enterprises based on the data, such as the industry to which the company belongs, the place of registration, the time of registration, and the scale of the company. In the user assistance information database, the user background information and the user equipment number are stored in one-to-one correspondence, so the extended user data corresponding to the user equipment number can be searched in the preset user assistance information database through the user equipment number.

After the extended user data and the user portrait data are obtained, the enhanced training data is constructed, and the expanded user data, user portrait data, original advertisement feature data and original conversion rate corresponding to each user equipment number are associated and saved, thereby Get augmented training data.

Step S20, using the original training data and the enhanced training data to perform joint model training to obtain an information conversion rate prediction model and an auxiliary model, and optimize the information conversion rate prediction model based on the auxiliary model;

In this embodiment, the original training data and the enhanced training data are used for joint model training. During the training process, the original training data and the enhanced training data are continuously used to optimize the initial prediction model and the auxiliary model based on the loss function. When converged, the information conversion rate prediction model is obtained.

Specifically, step S20 includes:

Step S21, input the original training data into the initial prediction model for model training, and input the enhanced training data into the auxiliary model for model training, based on the first prediction value of the initial prediction model and the first prediction value of the auxiliary model. Two predicted values, construct the first total loss function corresponding to the initial prediction model and the second total loss function of the auxiliary model;

Specifically, step S21 includes:

Step a, determining a first loss function based on the first predicted value of the initial prediction model and the original conversion rate of the original training data;

Step b, determining a third loss function based on the first predicted value of the initial prediction model and the second predicted value of the auxiliary model;

Step c, performing weighted summation on the first loss function and the third loss function to obtain the first total loss function.

Step d, determining a second loss function based on the second predicted value of the auxiliary model and the original conversion rate of the original training data;

Step e: Perform weighted summation on the third loss function and the second loss function determined based on the first predicted value of the initial prediction model and the second predicted value of the auxiliary model to obtain the second total loss function

In this embodiment, the loss function is used to estimate the degree of inconsistency between the predicted value of the model and the real value, and it is a non-negative real-valued function. The smaller the loss function, the better the robustness of the model. The process of training the model is to make the loss function smaller and smaller through continuous iterative calculation. The smaller the loss function, the better the algorithm is.

For the convenience of description, the original training data is represented by two-tuples:

<X _i ,Y _i >,(i=1,2,...,n)

Among them, X _i =(x ₁ , x ₂ ,...,x _m ) represents m features including user portrait data and original advertisement feature data; Y _i represents whether user behavior has been converted, 1 means yes, 0 means no .

Represent augmented training data as triples to improve the accuracy of predicting Y from data X:

<(X _i ,Z _i ),Y _i >,(i=1,2,...,n)

Among them, X _i =(x ₁ ,x ₂ ,...,x _m ) represents m features including user portrait data and original advertisement feature data; Z _i =(z ₁ ,z ₂ ,...,z _q ) represents q features including extended user data; Y _i represents whether the user behavior is transformed, 1 means yes, 0 means no.

Input the original training data into the initial prediction model f(X) for model training, and input the enhanced training data into the auxiliary model f'(X,Z) for model training, and use the first prediction value f(X _ij ) of the initial prediction model and the second predicted value f'(X _ij , Z _ij ) of the initial auxiliary model to construct a loss function.

During model training, the following optimization functions are used:

Among them, L ₁ is the loss function of the conversion rate estimation task, and L ₂ is the loss function of the auxiliary task. These two tasks are linked together by the L ₃ loss function. By solving the optimization function, a more accurate solution than the existing scheme is obtained. Conversion rate prediction model f(X).

Specifically, the initial prediction model and the auxiliary model are simultaneously trained, and the first total loss function corresponding to the initial prediction model and the second total loss function of the auxiliary model are respectively constructed. Based on the first predicted value of the initial prediction model and the original conversion rate of the original training data, a first loss function L ₁ (Y _ij -f(X _ij )) of the initial prediction model is determined, based on the first predicted value of the initial prediction model and The second predicted value of the initial auxiliary model determines the third loss function L ₃ (f'(X _ij , Z _ij ), f(X _ij )). The weighted summation of the first loss function and the third loss function is carried out to obtain the first total loss function of the initial prediction model:

L ₁ (Y _ij -f(X _ij ))+λL ₃ (f'(X _ij ,Z _ij ),f(X _ij ))

Among them, λ is an empirical value, and the value range is between 0 and 1, which is determined according to actual needs.

Similarly, for the second prediction value of the auxiliary model and the original conversion rate of the original training data, a second loss function L ₂ (Y _ij -f'(X _ij ,Z _ij )) is determined, based on the first The predicted value and the second predicted value of the initial auxiliary model determine a third loss function L ₃ (f'(X _ij , Z _ij ), f(X _ij )). The weighted summation of the second loss function and the third loss function is performed to obtain the second total loss function of the auxiliary model:

L ₂ (Y _ij -f(X _ij ))+λ'L ₃ (f'(X _ij ,Z _ij ),f(X _ij ))

Among them, λ' is an empirical value, and the value range is between 0 and 1, which is determined according to actual needs.

Step S22, performing model iterative training based on the first total loss function and the second total loss function, and using the second predicted value of the auxiliary model to optimize the initial prediction model;

In this embodiment, the first total loss function is used to train the initial prediction model, and the second total loss function is used to train the auxiliary model. When the model does not converge, the gradient information of the model is obtained by calculation, and then the model parameters of the initial prediction model and the model parameters of the auxiliary model are updated according to the gradient information, and the updated initial prediction model and auxiliary model are obtained. The auxiliary model continues with model training.

It should be noted that, in the model training process, the third loss function L ₃ (f'(X _ij , Z _ij ), f(X _ij )) is based on the first predicted value of the initial prediction model and the value of the auxiliary model. The second prediction value is determined, which means that the second prediction value of the auxiliary model is infinitely close to the first prediction value of the initial prediction model, so as to achieve the prediction effect that the information conversion rate prediction model is compatible with the auxiliary model, that is, the use of the auxiliary model is realized. The second prediction value of the model optimizes the initial prediction model.

Step S23, when the initial prediction model converges, the current initial prediction model is determined as the optimized information conversion rate prediction model.

Specifically, step S23 includes:

Step f, determining whether the initial prediction model converges based on the first total loss function and a preset threshold;

Step g: When the first total loss function is less than or equal to a preset threshold, it is determined that the initial prediction model is converged, and the optimized prediction model of the information conversion rate is obtained.

In this embodiment, in the iterative process of model training, after each round of training, the first total loss function value is compared with a preset threshold, and whether the model converges is determined according to the comparison result. When the model does not converge, iterative training needs to continue until the model converges and the model training stops. At this time, the initial prediction model obtained in this round of iteration is determined as the advertisement conversion rate prediction model. When the total loss function value is less than or equal to the preset threshold, the iteration is stopped. The preset threshold is determined according to the actual situation, and is generally a positive number close to 0. For example, the preset value may be 0.001. After multiple iterations, the first total loss function value will continue to decrease, and then approach the preset threshold. It is also possible to determine whether the model has converged according to the model training rounds. For example, the preset number of times can be 1000 or 20000 times. When the iteration reaches the preset number of times, the model stops training to obtain an advertisement conversion rate prediction model.

Step S30, when receiving the information conversion rate prediction request, obtain the conversion rate prediction result of the information content corresponding to the information conversion rate prediction request according to the information conversion rate prediction model.

In this embodiment, the advertisement conversion rate prediction request includes at least the user equipment number of the target user and the feature data of the advertisement to be placed. At this time, according to the user equipment number of the target user in the preset user portrait database, The device number is used to find the target user portrait data, and then the target user portrait data and the target advertisement feature data are input into the information conversion rate prediction model to predict the conversion rate, and the predicted value output by the information conversion rate prediction model is used as the predicted click conversion rate.

In the information conversion rate prediction method proposed in this embodiment, the extended user data corresponding to the user equipment number of the original training data is obtained, and the original training data is reconstructed based on the extended user data to obtain enhanced training data, which is then used The original training data and the enhanced training data are subjected to joint model training to obtain an information conversion rate prediction model, and then when an information conversion rate prediction request is received, the information conversion rate is obtained according to the information conversion rate prediction model. The conversion rate prediction result of the information content corresponding to the prediction request. By expanding user data to build enhanced training data, and using the original training data and enhanced training data for joint model training, a high-quality information conversion rate prediction model can be obtained, which improves the prediction accuracy of online information delivery, thereby improving the accuracy of information delivery. Success rate.

Based on the first embodiment, referring to FIG. 3 , a second embodiment of the information conversion rate prediction method of the present invention is proposed. In this embodiment, step S30 includes:

Step S31, in a preset user portrait database, search for target user portrait data according to the target user equipment number;

Step S32, inputting the target user portrait data and the target information feature data into the optimized information conversion rate prediction model for conversion rate prediction, and using the predicted value output by the information conversion rate prediction model as the predicted conversion Rate.

In this embodiment, the enhanced training data is constructed by expanding the user data, and the original training data and the enhanced training data are used for joint model training, so as to obtain a high-quality information conversion rate prediction model, and then the information conversion rate prediction model is used to determine the The precise target group corresponding to the advertisement placed.

The advertisement conversion rate prediction request includes at least the user device number of the target user and the feature data of the advertisement to be placed. At this time, in the preset user portrait database, the target user portrait data is searched according to the target user device number, and then the target user portrait is searched. The data and target advertisement feature data are input into the information conversion rate prediction model to predict the conversion rate. The predicted value output by the information conversion rate prediction model is the conversion rate. When the predicted value is 1, it means that the user corresponding to the target user equipment number is the The precise target group corresponding to the advertisement to be placed, otherwise, when the predicted value is 0, it means that the user corresponding to the user equipment number is not interested in the advertisement, and the possibility of purchasing the advertised product is extremely low. Therefore, according to the prediction results of the information conversion rate prediction model, online advertisements can be guided to accurate target users, and the success rate of accurate advertisements can be improved.

In the information conversion rate prediction method proposed in this embodiment, in the preset user portrait database, the target user portrait data is searched according to the target user equipment number, and then the target user portrait data and the target information feature data are input into the desired user portrait data. The information conversion rate prediction model is used to predict the conversion rate, and the predicted value output by the information conversion rate prediction model is called the predicted conversion rate. Use the original training data and the enhanced training data for joint model training, obtain the information conversion rate prediction model, and predict the click conversion rate, which improves the prediction accuracy of online delivery, thereby improving the success rate of accurate information delivery

The present invention further provides an apparatus for predicting information conversion rate. Referring to FIG. 4 , FIG. 4 is a schematic diagram of functional modules of an embodiment of the apparatus for predicting information conversion rate according to the present invention.

The building module 10 is used to obtain the extended user data corresponding to the user equipment number of the original training data, and reconstruct the original training data based on the extended user data to obtain enhanced training data;

The training module 20 is used to perform joint model training using the original training data and the enhanced training data to obtain an information conversion rate prediction model and an auxiliary model, and optimize the information conversion rate prediction model based on the auxiliary model ;

The prediction module 30 is configured to obtain, when an information conversion rate prediction request is received, a conversion rate prediction result of the information content corresponding to the information conversion rate prediction request according to the optimized information conversion rate prediction model.

Further, the building block 10 is also used for:

In the preset user assistance information database, look up the extended user data according to the user equipment number, and in the preset user portrait database, look up the user portrait data according to the user equipment number;

Based on the user equipment number, the extended user data, the user portrait data, the original information feature data, and the original conversion rate are associated and saved as the enhanced training data.

Further, the training module 20 is also used for:

Input the original training data into the initial prediction model for model training, while inputting the enhanced training data into the auxiliary model for model training, based on the first prediction value of the initial prediction model and the second prediction value of the auxiliary model , construct the first total loss function corresponding to the initial prediction model and the second total loss function of the auxiliary model;

Perform model iterative training based on the first total loss function and the second total loss function, and use the second predicted value of the auxiliary model to optimize the initial prediction model;

When the initial prediction model converges, the current initial prediction model is determined as the optimized information conversion rate prediction model.

Further, the training module 20 is also used for:

determining a first loss function based on the first predicted value of the initial prediction model and the original conversion rate of the original training data;

determining a third loss function based on the first predicted value of the initial prediction model and the second predicted value of the auxiliary model;

Weighted summation is performed on the first loss function and the third loss function to obtain the first total loss function.

Further, the training module 20 is also used for:

determining whether the initial prediction model converges based on the first total loss function and a preset threshold;

When the first total loss function is less than or equal to a preset threshold, it is determined that the initial prediction model is converged, and the optimized prediction model of the information conversion rate is obtained.

Further, the training module 20 is also used for:

determining a second loss function based on the second predicted value of the auxiliary model and the original conversion rate of the original training data;

A weighted summation is performed on the third loss function and the second loss function determined based on the first predicted value of the initial prediction model and the second predicted value of the auxiliary model, to obtain the second total loss function.

Further, the prediction module 30 is also used for:

In the preset user portrait database, search for target user portrait data according to the target user equipment number;

The target user portrait data and the target information feature data are input into the information conversion rate prediction model for conversion rate prediction, and the predicted value output by the information conversion rate prediction model is called the predicted conversion rate.

In addition, an embodiment of the present invention further provides a readable storage medium, where an information conversion rate prediction program is stored on the readable storage medium, and when the information conversion rate prediction program is executed by a processor, the information conversion in the above-mentioned embodiments is realized The steps of the rate prediction method.

It should be noted that, herein, the terms "comprising", "comprising" or any other variation thereof are intended to encompass non-exclusive inclusion, such that a process, method, article or system comprising a series of elements includes not only those elements, It also includes other elements not expressly listed or inherent to such a process, method, article or system. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in the process, method, article or system that includes the element.

The above-mentioned serial numbers of the embodiments of the present invention are only for description, and do not represent the advantages or disadvantages of the embodiments.

From the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus a necessary general hardware platform, and of course hardware can also be used, but in many cases the former is better implementation. Based on such understanding, the technical solutions of the present invention can be embodied in the form of software products in essence or the parts that make contributions to the prior art, and the computer software products are stored in a storage medium (such as ROM/RAM) as described above. , magnetic disk, optical disk), including several instructions to make a system device (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in the various embodiments of the present invention.

The above are only preferred embodiments of the present invention, and are not intended to limit the scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the description and drawings of the present invention, or directly or indirectly applied in other related technical fields , are similarly included in the scope of patent protection of the present invention.

Claims (10)

1. An information conversion rate prediction method, characterized by comprising:

acquiring extended user data corresponding to a user equipment number of original training data, and reconstructing the original training data based on the extended user data to obtain enhanced training data;

performing combined model training by using the original training data and the enhanced training data to obtain an information conversion rate prediction model and an auxiliary model, and optimizing the information conversion rate prediction model based on the auxiliary model;

and when an information conversion rate prediction request is received, obtaining a conversion rate prediction result of the information content corresponding to the information conversion rate prediction request according to the optimized information conversion rate prediction model.

2. The information conversion rate prediction method according to claim 1, wherein the original training data includes original information characteristic data and an original conversion rate corresponding to a user equipment number;

the step of obtaining extended user data corresponding to the user equipment number of the original training data, and reconstructing the original training data based on the extended user data to obtain enhanced training data includes:

searching the expanded user data according to the user equipment number in a preset user auxiliary information base, and searching user portrait data according to the user equipment number in a preset user portrait database;

and storing the extended user data, the user portrait data, the original information characteristic data and the original conversion rate association as the enhanced training data based on the user equipment number.

3. The information conversion rate prediction method according to claim 1, wherein the step of performing joint model training using the original training data and the enhanced training data to obtain an information conversion rate prediction model and an auxiliary model, and optimizing the information conversion rate prediction model based on the auxiliary model comprises:

inputting the original training data into an initial prediction model for model training, inputting the enhanced training data into an auxiliary model for model training, and constructing a first total loss function corresponding to the initial prediction model and a second total loss function corresponding to the auxiliary model based on a first predicted value of the initial prediction model and a second predicted value of the auxiliary model;

performing model iterative training based on the first total loss function and the second total loss function, and optimizing the initial prediction model by using a second prediction value of the auxiliary model;

and when the initial prediction model converges, determining the current initial prediction model as the optimized information conversion rate prediction model.

4. The information conversion rate prediction method according to claim 3, wherein the step of constructing a first total loss function corresponding to the initial prediction model based on the first predicted value of the initial prediction model and the second predicted value of the auxiliary model comprises:

determining a first loss function based on a first predicted value of the initial prediction model and an original conversion rate of the original training data;

determining a third loss function based on the first predicted value of the initial prediction model and the second predicted value of the auxiliary model;

and carrying out weighted summation on the first loss function and the third loss function to obtain the first total loss function.

5. The information conversion rate prediction method according to claim 3, wherein the step of determining the current initial prediction model as the optimized information conversion rate prediction model when the initial prediction model converges comprises:

determining whether the initial prediction model converges based on the first total loss function and a preset threshold;

and when the first total loss function is smaller than or equal to a preset threshold value, determining the convergence of the initial prediction model to obtain the optimized information conversion rate prediction model.

6. The information conversion rate prediction method according to claim 3, wherein the step of constructing a second total loss function corresponding to the auxiliary model based on the first predicted value of the initial prediction model and the second predicted value of the auxiliary model comprises:

determining a second loss function based on a second predicted value of the auxiliary model and an original conversion rate of the original training data;

and performing weighted summation on a third loss function and the second loss function determined based on the first predicted value of the initial prediction model and the second predicted value of the auxiliary model to obtain a second total loss function.

7. The information conversion rate prediction method according to any one of claims 1 to 6, wherein the information conversion rate prediction request includes a target user equipment number and target information characteristic data, and the step of obtaining, when receiving the information conversion rate prediction request, a conversion rate prediction result of information content corresponding to the information conversion rate prediction request according to the optimized information conversion rate prediction model includes:

searching target user portrait data in a preset user portrait database according to the target user equipment number;

inputting the target user portrait data and the target information characteristic data into the optimized information conversion rate prediction model for conversion rate prediction, and taking a prediction value output by the information conversion rate prediction model as the prediction conversion rate.

8. An information conversion rate prediction apparatus, characterized by comprising:

the construction module is used for acquiring expanded user data corresponding to the user equipment number of the original training data and reconstructing the original training data based on the expanded user data to obtain enhanced training data;

the training module is used for performing combined model training by using the original training data and the enhanced training data to obtain an information conversion rate prediction model and an auxiliary model, and optimizing the information conversion rate prediction model based on the auxiliary model;

and the prediction module is used for obtaining a conversion rate prediction result of the information content corresponding to the information conversion rate prediction request according to the optimized information conversion rate prediction model when the information conversion rate prediction request is received.

9. An information conversion rate prediction apparatus characterized by comprising: a memory, a processor, and an information conversion rate prediction program stored on the memory and executable on the processor, the information conversion rate prediction program when executed by the processor implementing the steps of the information conversion rate prediction method according to any one of claims 1 to 7.

10. A readable storage medium, characterized in that the readable storage medium has stored thereon an information conversion rate prediction program which, when executed by a processor, realizes the steps of the information conversion rate prediction method according to any one of claims 1 to 7.

Images