CN118656730B

CN118656730B - Financing scheme generation method and system based on supply chain data

Info

Publication number: CN118656730B
Application number: CN202410794924.7A
Authority: CN
Inventors: 方益
Original assignee: Minxing Information Technology Co ltd
Current assignee: Minxing Information Technology Co ltd
Priority date: 2024-06-19
Filing date: 2024-06-19
Publication date: 2024-11-22
Anticipated expiration: 2044-06-19
Also published as: CN118656730A

Abstract

The invention provides a financing scheme generation method and system based on supply chain data, which relate to the technical field of financial services and comprise the steps of acquiring first information and second information; performing evaluation processing according to the first information and the second information to obtain a supply chain influence scoring result; performing data fusion processing on the second information to obtain a supply chain fusion data set; performing identification processing according to the supply chain fusion data set to obtain a supply chain network model; carrying out data integration processing on the enterprise basic data, the supply chain influence scoring result and the supply chain network model based on the graphic neural network mathematical model to obtain a comprehensive data view; performing time sequence analysis processing according to the comprehensive data view to obtain an evaluation result; and optimizing the evaluation result based on the deep learning mathematical model to obtain a personalized financing scheme. The multi-source data fusion mathematical model based on Bayesian reasoning can effectively process multi-source heterogeneous data in a supply chain, and realize efficient fusion of the data.

Description

Financing scheme generation method and system based on supply chain data

Technical Field

The invention relates to the technical field of financial services, in particular to a financing scheme generation method and system based on supply chain data.

Background

With the continued development of global economies and the increasing complexity of supply chain management, the importance of financial services for supply chain-related enterprises in modern economies is becoming increasingly prominent. The supply chain financial service aims to provide a personalized financing scheme for enterprises in the supply chain through the close cooperation of financial institutions and enterprises in various links of the supply chain, so that the fund liquidity and the overall operation efficiency are improved. However, current supply chain financial services face a number of challenges and problems in practical applications. Firstly, the existing financing scheme making mode mainly depends on credit assessment of a single enterprise, and the method only analyzes based on financial data and credit records provided by the enterprise, ignores complex relations and dynamic changes of the enterprise in a supply chain, and makes a risk assessment result inaccurate. Secondly, the prior art is relatively backward in terms of data processing, and is generally analyzed by adopting a simple statistical and linear regression model, so that the effective fusion and deep analysis of multi-source heterogeneous data are lacked, the potential value of large data is difficult to fully mine and utilize, and the deviation of an evaluation result and the unreasonable financing scheme are caused.

Based on the above-mentioned drawbacks of the prior art, there is a need for a financing scheme generation method and system based on supply chain data.

Disclosure of Invention

The invention aims to provide a financing scheme generation method and system based on supply chain data so as to solve the problems. In order to achieve the above purpose, the technical scheme adopted by the invention is as follows:

In a first aspect, the present application provides a method for generating a financing scheme based on supply chain data, comprising:

Acquiring first information and second information, wherein the first information comprises enterprise basic data for applying for financial services, and the second information comprises enterprise basic information in a supply chain, supply chain transaction data, logistics information and financial institution data;

performing evaluation processing according to the first information and the second information, and obtaining a supply chain influence scoring result of the enterprise by evaluating the importance degree and the replaceability of the enterprise in the supply chain;

performing data fusion processing on the second information by using a multi-source data fusion mathematical model based on Bayesian reasoning to obtain a supply chain fusion data set;

Performing identification processing according to the supply chain fusion data set, identifying key node enterprises in a supply chain through social network analysis, and evaluating the centrality and connectivity of the key node enterprises in the whole supply chain to obtain a supply chain network model;

Performing data integration and association mining processing on the enterprise basic data, the supply chain influence scoring result and the supply chain network model based on a preset graphic neural network mathematical model to obtain a comprehensive data view, wherein the comprehensive data view comprises network positions and relationship strengths of enterprises applying financial services in a supply chain;

Performing time sequence analysis processing according to the comprehensive data view to obtain an evaluation result, wherein the evaluation result comprises a credit evaluation result, a risk prediction result and a systematic risk score in a supply chain of the enterprise;

And carrying out multi-objective optimization processing on the evaluation result based on a preset deep learning mathematical model, generating at least two financing schemes under different scenes, and selecting and adjusting the financing schemes through a hybrid intelligent algorithm to obtain a final personalized financing scheme.

In a second aspect, the present application also provides a financing scheme generation system based on supply chain data, comprising:

The system comprises an acquisition module, a data processing module and a data processing module, wherein the acquisition module is used for acquiring first information and second information, the first information comprises enterprise basic data for applying for financial services, and the second information comprises enterprise basic information in a supply chain, supply chain transaction data, logistics information and financial institution data;

The evaluation module is used for performing evaluation processing according to the first information and the second information, and obtaining a supply chain influence scoring result of the enterprise by evaluating the importance degree and the replaceability of the enterprise in the supply chain;

the fusion module is used for carrying out data fusion processing on the second information based on a multi-source data fusion mathematical model of Bayesian reasoning to obtain a supply chain fusion data set;

The identification module is used for carrying out identification processing according to the supply chain fusion data set, identifying key node enterprises in a supply chain through social network analysis, and evaluating the centrality and connectivity of the key node enterprises in the whole supply chain to obtain a supply chain network model;

The integration module is used for carrying out data integration and association mining processing on the enterprise basic data, the supply chain influence scoring result and the supply chain network model based on a preset graph neural network mathematical model to obtain a comprehensive data view, wherein the comprehensive data view comprises network positions and relationship strengths of enterprises applying financial services in a supply chain;

The analysis module is used for carrying out time sequence analysis processing according to the comprehensive data view to obtain an evaluation result, wherein the evaluation result comprises a credit evaluation result, a risk prediction result and a systematic risk score in a supply chain of the enterprise;

And the optimizing module is used for carrying out multi-objective optimizing processing on the evaluation result based on a preset deep learning mathematical model, generating at least two financing schemes under different scenes, and selecting and adjusting the financing schemes through a hybrid intelligent algorithm to obtain a final personalized financing scheme.

The beneficial effects of the invention are as follows:

the multi-source data fusion mathematical model based on Bayesian reasoning can effectively process multi-source heterogeneous data in a supply chain, and realize efficient fusion of the data. The complex relationship among supply chain enterprises is accurately reflected through clustering, condition summarization calculation and maximum posterior estimation, and a more accurate supply chain fusion data set is obtained.

Additional features and advantages of the invention will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the embodiments of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims thereof as well as the appended drawings.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic flow chart of a financing scheme generation method based on supply chain data according to an embodiment of the present invention;

FIG. 2 is a flow chart of a genetic algorithm described in an embodiment of the present invention.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

It should be noted that: like reference numerals and letters denote like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures. Meanwhile, in the description of the present invention, the terms "first", "second", and the like are used only to distinguish the description, and are not to be construed as indicating or implying relative importance.

Example 1:

the embodiment provides a financing scheme generation method based on supply chain data.

Referring to fig. 1, the method is shown to include steps S100, S200, S300, S400, S500, S600, and S700.

Step S100, acquiring first information and second information, wherein the first information comprises enterprise basic information for applying for financial services, and the second information comprises enterprise basic information in a supply chain, supply chain transaction data, logistics information and financial institution data;

It will be appreciated that the business base includes primarily the basic information of the business applying for financial services, such as business name, registered capital, legal representatives, camping service, historical operating data, financial statements (liability statement, profit statement, cash flow statement), credit records, stakeholders, etc. Obtaining such data typically requires the enterprise to actively submit and verify and supplement through an enterprise internal information system or a third party data provider. The business base information in the supply chain includes base information of upstream suppliers and downstream customers related to the application business, such as business names, contacts, business collaboration history, etc. Supply chain transaction data relates to transaction records, including purchase orders, sales orders, invoice information, payment records, and the like, between an enterprise and its supply chain upstream and downstream partners, reflecting the transaction behavior and credit status of the enterprise in the supply chain. The logistics information comprises data such as a transportation path, a transportation mode, storage information, a logistics provider and the like of goods, and the data are used for analyzing the fluidity and logistics efficiency of a supply chain. The financial institution data comprises credit records, loan interest rates, credit approval standards, repayment records and the like of enterprises in the supply chain, and the data can be acquired from banks, financial service platforms and the like to reflect financing history and credit conditions of the enterprises. In the data acquisition process, data privacy protection laws and regulations are strictly complied with, security of enterprise sensitive information is ensured, and data leakage and abuse are prevented.

Step 200, performing evaluation processing according to the first information and the second information, and obtaining a supply chain influence scoring result of the enterprise by evaluating the importance degree and the replaceability of the enterprise in the supply chain;

It can be understood that the status and the importance degree of the enterprise in the supply chain are comprehensively evaluated by analyzing factors such as the transaction frequency, the transaction amount, the transaction stability, the logistics path and the like of the enterprise in the supply chain; the degree of replaceability of the enterprise is assessed by analyzing information such as trade relation, supply chain network structure, logistics path and the like of the enterprise. The supply chain influence scoring result is a comprehensive index reflecting the status and influence degree of the enterprise in the supply chain. By evaluating the influence of the supply chain of the enterprise, key nodes and cooperation opportunities in the supply chain can be found, cooperation and optimization of the supply chain are promoted, and overall efficiency and competitiveness are improved.

Step S300, performing data fusion processing on the second information by using a multi-source data fusion mathematical model based on Bayesian reasoning to obtain a supply chain fusion data set;

It will be appreciated that the data sources involved in the supply chain typically come from different sources and different formats, and that there may be instances of inconsistent, incomplete, and inaccurate data. Therefore, it is necessary to integrate and uniformly process these data by means of a data fusion technique to eliminate conflicts and contradictions between the data, thereby obtaining a more reliable data set. Bayesian reasoning is a probability inference method based on bayesian theorem, and the posterior probability is obtained by continuously updating the prior probability according to new evidence. In data fusion, bayesian reasoning is utilized to handle uncertainty and inconsistency between different data sources, thereby better fusing data. In the step, bayesian reasoning and data fusion processing are carried out on the basic information of enterprises, the transaction data of the supply chain, the logistics information and the financial institution data, so that a comprehensive supply chain fusion data set can be obtained. This data set will contain more comprehensive, more accurate supply chain information, providing more reliable data support for subsequent analysis and decision making.

Step S400, carrying out identification processing according to a supply chain fusion data set, identifying key node enterprises in a supply chain through social network analysis, and evaluating the centrality and connectivity of the key node enterprises in the whole supply chain to obtain a supply chain network model;

Further, the constructed supply chain network model can reflect the relation among enterprises in the supply chain more comprehensively and accurately, and is beneficial to the financial institutions to know the structure and the operation mode of the supply chain comprehensively. Centrality is an index for measuring the importance degree of nodes in a network, and common centrality indexes comprise centrality, medium centrality and near centrality. By centrally evaluating key node enterprises, one can learn their impact and position in the supply chain to better understand and manage the supply chain network. Based on the identification and centrality evaluation results of the key node enterprises, a supply chain network model can be constructed. This model will show the relationships between the various enterprises in the supply chain, helping the financial institutions to fully understand the interactive relationships between the supply chain structure and the enterprises.

Step S500, carrying out data integration and association mining processing on enterprise basic data, a supply chain influence scoring result and a supply chain network model based on a preset graph neural network mathematical model to obtain a comprehensive data view, wherein the comprehensive data view comprises network positions and relationship strengths of enterprises applying financial services in a supply chain;

The graph neural network model can fully utilize structural information and node characteristics of graph data, so that effective representation and association mining of the graph structure are realized, potential association and rules in the data can be mined, potential cooperation relations, influence factors and possible risk points among enterprises can be found, and deeper data insight is provided for financial institutions.

Step S600, performing time sequence analysis processing according to the comprehensive data view to obtain an evaluation result, wherein the evaluation result comprises a credit evaluation result, a risk prediction result and a systematic risk score in a supply chain of the enterprise;

It will be appreciated that by time series analysis, we will analyze and compare the data of the business at different points in time to reveal the evolving trend of its credits and risks and make predictions and evaluations for the future. Taking into account the complexity and interdependence of the supply chain, the risk assessment of the enterprise is incorporated into the perspective of the entire supply chain, helping the financial institution to better understand and assess the stability and risk distribution of the entire supply chain network.

And step S700, performing multi-objective optimization processing on the evaluation result based on a preset deep learning mathematical model, generating at least two financing schemes under different scenes, and selecting and adjusting the financing schemes through a hybrid intelligent algorithm to obtain a final personalized financing scheme.

First, using a deep learning mathematical model, multi-objective optimization processing is performed for different objectives (e.g., financing cost, risk minimization, liquidity maximization, etc.), ensuring that the generated financing scheme is advantageous in multiple aspects. Based on the results of the multi-objective optimization process, at least two financing schemes in different scenes are generated, wherein the schemes comprise different interest rates, deadlines, repayment modes and the like so as to meet the requirements and conditions of different enterprises. The generated financing schemes are selected and adjusted by mixed intelligent algorithms (such as genetic algorithms, simulated annealing algorithms and the like), and the algorithms can find the optimal personalized financing scheme while guaranteeing the diversity of the schemes.

The step S200 includes a step S210, a step S220, a step S230, and a step S240.

Step S210, importance evaluation processing is carried out according to enterprise basic data and supply chain transaction data, and importance scores of enterprises applying for financial services in a supply chain are obtained by constructing a transaction network diagram and calculating PageRank values of each enterprise by adopting a PageRank algorithm;

First, a transaction network diagram is constructed based on supply chain transaction data. In this figure, each node represents one enterprise, and each edge represents a trade relationship between two enterprises, so that the trade relationship between enterprises in the supply chain can be intuitively displayed. And then, calculating the constructed transaction network graph by adopting a PageRank algorithm, and applying the PageRank algorithm to enterprise nodes in a supply chain to determine the importance degree of each enterprise in the supply chain, wherein the PageRank algorithm can comprehensively consider the direct transaction relationship of the enterprise and the connection condition with other important enterprises, so that the importance of the supply chain of the enterprise can be more comprehensively evaluated. Finally, the PageRank value of each enterprise, namely the importance score in the supply chain, can be obtained through the calculation result of the PageRank algorithm. This score reflects the status and impact of the business in the supply chain and can be an important basis for subsequent assessment.

Step S220, performing redundant network analysis according to enterprise basic data and logistics information, and obtaining an alternative score by calculating the redundancy degree of each node enterprise in a supply chain and evaluating the position and influence of an application financial service enterprise in a logistics path;

Specifically, it is first necessary to collect comprehensive enterprise base data including information such as company profiles, operation histories, and key performance indicators. Such data may provide insight regarding the structure and characteristics of each enterprise in the supply chain. Next, logistical information is collected, including data related to the movement of goods and materials in the supply chain. This includes detailed information about the route of transportation, time of delivery, inventory levels, and potential bottlenecks or breaks in the logistics network. The redundancy level of each node enterprise is then evaluated, which involves analyzing factors such as the number of available alternative routes, the degree of overlap in functionality and capability between different enterprises, and the toughness of the supply chain network to interruptions. Based on this analysis, each enterprise is assigned a redundancy score indicating its degree of redundancy in the supply chain. In addition, the location and impact of the applicant company in the logistics path is assessed to determine its alternatives or other options in the event of a break or change in the supply chain.

Step S230, carrying out weighted comprehensive processing according to the importance scores and the substitutability scores to obtain a preliminary scoring result;

In this step, the importance score and the alternative score are weighted first. Such weighting may be based on the importance of different factors, such as strategic status, market share, or business relevance of each node enterprise in the supply chain. These weights can be adjusted and determined based on actual conditions and domain expertise. The importance scores and the alternative scores for each business are then weighted together to obtain their overall scores in the supply chain. The weighted summation method may be a simple weighted average or may be calculated according to a specific weighting function to better reflect the relationship between different factors. Finally, through the weighted comprehensive processing, a preliminary scoring result of each enterprise can be obtained. These preliminary scoring results will provide an important reference for subsequent analysis and decision making, helping to better understand the status and impact of the application business in the supply chain, thereby providing more accurate and personalized financial services thereto.

And step 240, performing adjustment processing according to the preliminary scoring result and the enterprise historical credit record in the financial institution data, and performing calibration processing on the preliminary scoring result by using a weighted linear regression model to obtain a final supply chain influence scoring result.

Specifically, the enterprise history credit record in the financial institution data is utilized first, including information such as credit rating, loan record, repayment record, etc. of the enterprise. A weighted linear regression model is then created that uses the preliminary scoring results as independent variables and the enterprise historical credit records in the financial institution data as dependent variables to fit the relationship between the two by regression analysis. Weights of different credit records are taken into account when modeling, as well as non-linear relationships that may exist. And then, using the established weighted linear regression model, the preliminary scoring results can be calibrated to obtain more accurate and reliable supply chain influence scoring results. The calibration processing can be combined with enterprise historical credit records in financial institution data on the basis of the preliminary scoring result, so that the accuracy and objectivity of scoring are further improved. The supply chain influence scoring result obtained through adjustment processing provides important references for financial institutions, helps the financial institutions to better know the status and credit conditions of enterprises in the supply chain, and accordingly financial service strategies are formulated more accurately, risks are reduced, and efficiency is improved.

The step S300 includes a step S310, a step S320, a step S330, and a step S340.

Step S310, clustering is carried out according to supply chain transaction data and logistics information, transaction amount, transaction frequency and logistics node information in a supply chain are processed by using a Gaussian mixture model, and multidimensional normal distribution of the data is fitted to obtain initial multidimensional data distribution;

It will be appreciated that the supply chain transaction data and logistics information include the amount of transactions, frequency of transactions, and location and characteristics of logistics nodes between different businesses, etc. These data are typically multidimensional and contain information in different ways. The data is then processed using a cluster analysis method to identify potential patterns and structures therein. In order to better describe the distribution situation of the data, a Gaussian mixture model is used for fitting the multidimensional normal distribution of the data, and the distribution situation of the data is approximately represented by a group of Gaussian distributions, so that the characteristics and the structure of the data are better understood, and an important basis is provided for subsequent data fusion and processing.

Step S320, carrying out condition summarization calculation according to initial multidimensional data distribution and financial institution data, classifying loan interest rates, credit approval standards and repayment records by using a naive Bayesian classifier, and calculating posterior probabilities of all categories to obtain posterior probability sets of all categories;

In particular, financial institution data includes loan interest rates, credit approval criteria, and repayment records, etc., which will be used to aid in understanding and assessing the loan risk and credit status of the business. Next, a method of conditional generalization computation will be employed, which is one of the key steps of the naive bayes classifier. By conditional summary calculation, the conditional probability distribution of other variables under given conditions will be calculated based on known categories (e.g., loan interest rate, credit approval criteria, and repayment records, etc.). A naive bayes classifier, a probabilistic classification algorithm, will then be applied to deal with the classification problem. In this embodiment, the naive bayes classifier is used to classify the enterprises into different categories, such as "high risk", "medium risk" and "low risk", by using factors such as loan interest rate, credit approval standard and repayment record of the enterprises as features. Finally, the posterior probabilities for each class will be calculated, which is an estimate of the probabilities for the different classes given the observed data. By calculating the posterior probability set, probability distribution conditions of enterprises belonging to different risk categories can be obtained, so that credit conditions and loan risks of the enterprises can be better evaluated. The method comprises the steps of classifying factors such as loan interest rate, credit approval standard, repayment record and the like of enterprises through condition summarization calculation and a naive Bayesian classifier, and calculating posterior probability of each category, so that credit condition and loan risk of the enterprises are evaluated more accurately.

Step S330, performing maximum posterior estimation processing according to the posterior probability set, and constructing a causal relationship graph by utilizing a Bayesian network to obtain a fusion causal network reflecting the relevance between enterprises of the supply chain;

Specifically, the maximum a posteriori estimation processing method is a method of estimating parameters by taking a priori knowledge and observation data into consideration. In this step, the association between supply chain enterprises is estimated by maximum posterior estimation using a posterior probability set as observation data, in combination with possibly existing prior knowledge. Next, a causal relationship graph will be constructed using Bayesian networks. A bayesian network is a probabilistic graph model representing the dependency between variables. In this embodiment, a causal relationship graph between supply chain enterprises will be established through a bayesian network, where nodes represent enterprises and edges represent relationships between enterprises, such as partnerships, trade relationships, and the like. By constructing a bayesian network, the associations and interrelationships between supply chain enterprises can be better understood and described. Such a converged causal network may help identify key node enterprises in the supply chain, thereby better assessing the stability, reliability, and risk of the supply chain.

And step 340, sampling and deducing the fusion causal network based on a preset Markov chain Monte Carlo mathematical model to obtain a supply chain fusion data set.

It will be appreciated that Markov Chain Monte Carlo (MCMC) is a statistical simulation method for randomly sampling from a probability distribution. In the step, a Markov chain Monte Carlo method is utilized to extract samples from joint probability distribution of a fusion causal network so as to acquire relevance and interaction information among supply chain enterprises. Then, the sampled sample data is subjected to an estimation process. By carrying out statistical analysis and inference on the sampled samples, probability distribution of various indexes among enterprises of the supply chain, such as relationship strength, influence degree and the like among the enterprises, can be obtained. These inferences will help to understand the nature and structure of the supply chain more deeply. The final generated supply chain fusion data set contains the relevance and interaction information among the supply chain enterprises, and provides an important data basis for subsequent analysis and application.

The step S400 includes a step S410, a step S420, a step S430, and a step S440.

Step S410, performing K core decomposition processing according to the supply chain fusion data set, and obtaining a preliminary key node set by removing low-level nodes;

It can be understood that the K-core decomposition algorithm is a graph theory algorithm for identifying key nodes in the graph, and the algorithm progressively strips non-key nodes in the graph by continuously removing nodes with degrees lower than K and their related edges, so as to finally obtain a group of key nodes with higher degrees and close connection. In the step, K core decomposition processing is carried out on the supply chain fusion data set according to the set K value. Nodes with degrees lower than K are removed continuously and iteratively, and nodes with small influence on the overall structure of the supply chain are gradually removed, so that a group of preliminary key nodes are screened out. The K core decomposition treatment is carried out to obtain a preliminary key node set of the supply chain, the nodes have higher degree and tight connectivity, play an important role in the whole supply chain network, have important influence on the stability and efficiency of the supply chain, and are used as the basis of the subsequent social network analysis to help identify key node enterprises in the supply chain.

Step S420, calculating the times of each node serving as a supply chain intermediate node according to the preliminary key node set to obtain a betweenness centrality score;

it should be noted that the betting center is a common index in social network analysis, and is used to measure the betting degree of a node playing in a network. In a supply chain network, the betting center may reflect the importance of one node in the path connecting different nodes. If a node appears in the shortest path between many node pairs, its median centrality is high. The calculation formula is as follows:

Where C _B (v) is a node v betweenness centrality score, s, t and v are node names, σ _st represents the total number of shortest paths from node s to node t, and σ _st (v) represents the number of paths through node v among all the shortest paths from node s to node t. In calculating the betting centrality, we will traverse all node pairs s and t and calculate the shortest path number through node v. Then, the number of shortest paths passing through the node v is divided by the total number of shortest paths to obtain a median centrality score of the node v. The step can accurately evaluate the intermediation degree of each node in the supply chain network by calculating the intermediation centrality. This helps identify nodes that play a critical mediating role in the supply chain, providing an important basis for subsequent supply chain network model construction and analysis.

Step S430, calculating the degree and distance of each node according to the supply chain fusion data set and the betweenness centrality score to obtain a centrality score and a near centrality score of the node in the supply chain;

in calculating the centrality and the near centrality, the degree of direct connection of the nodes in the network and the distance of the nodes from other nodes need to be considered. These metrics may help understand the importance and location of nodes in the supply chain. The centrality reflects the direct connection condition of the nodes, and the proximity centrality considers the distance relation between the nodes and other nodes. By calculating the centrality and approaching centrality, the importance and location of each node in the supply chain network can be more fully assessed. Wherein, the centrality score calculation formula:

the approximate centrality score calculation formula:

Where V, u are node names, C _D (V) represents the degree centrality score of node V, deg (V) represents the degree of node V, i.e. the number of directly connected nodes, C _C (V) the proximity centrality score of node V, N represents the total number of nodes in the network, d (V, u) represents the shortest path distance from node V to node u, and V represents the set of nodes in the network.

And S440, performing weighted clustering coefficient calculation according to the betweenness centrality score, the degree centrality score and the proximity centrality score, obtaining the final centrality and connectivity of the key node enterprise by calculating the local network compactness of the nodes, and integrating the final centrality and connectivity with the supply chain fusion data set to obtain the supply chain network model.

It will be appreciated that the betting centrality score can help determine which nodes play a key role in information dissemination and interaction, the centrality score then reflects the direct connection of the nodes, and the proximity centrality score takes into account the distance of the nodes from other nodes. By comprehensively considering the indexes, key node enterprises in the supply chain can be identified more accurately, the enterprises play a core role in the whole supply chain network, and an important supporting role is played for the stability and operation of the whole supply chain. Compared with simple centrality or betweenness centrality evaluation, the weighted clustering coefficient calculation method can better reflect the real situation of the supply chain network, thereby providing more reliable data support for subsequent risk evaluation and financing scheme formulation. In addition, the supply chain network model is closer to the actual situation by integrating the supply chain fusion data set, so that the accuracy and the credibility of the evaluation result are improved.

The step S500 includes a step S510, a step S520, a step S530, and a step S540.

Step S510, carrying out data integration and association mining processing on enterprise basic data, a supply chain influence scoring result and a supply chain network model based on a preset graphic neural network mathematical model to obtain an association information representation;

The key of this step is to integrate different types of data by using the capability of the graph neural network and mine the association relationship between them. In this way, the location, impact, and relationship with other businesses in the supply chain may be more fully understood. First, the graph neural network is capable of processing data at the node level and the graph level. In this embodiment, the enterprise's underlying data, supply chain impact scoring results, and supply chain network models need to be integrated and interrelated mined, meaning that the characteristics of individual enterprises and their location and impact throughout the supply chain network need to be considered simultaneously. The graph neural network is able to effectively capture complex correlations between these data and convert them into a representation of features that can be used for subsequent processing. Second, the graph neural network is capable of processing different types of graph structure data, including directed graphs, undirected graphs, and graphs with node and edge attributes. Also in this embodiment, what is needed is a complex supply chain network that includes both relationships between businesses and their attribute information, such as business infrastructure and supply chain transaction data. The flexibility of the graph neural network allows these different types of data to be integrated together to construct a comprehensive graph model to better understand the relationships and features between the various node enterprises in the supply chain. Finally, the graphic neural network has learning and reasoning capabilities, can learn complex modes and rules from data and is used for subsequent prediction and decision, and the graphic neural network is utilized to learn association relations among all node enterprises in a supply chain from basic data, supply chain influence scoring results and a supply chain network model of the enterprises so as to predict credit evaluation results, risk prediction results and systematic risk scoring of the enterprises. The end-to-end learning mode can better mine information in data, and provides more accurate and reliable decision support for supply chain financial services.

Step S520, carrying out community division processing according to the association information representation, and dividing enterprises closely associated with each other in the graph into the same community to obtain a community division result, wherein the community division result comprises network positions of the enterprises in a supply chain;

It will be appreciated that complex associations exist between enterprises in the supply chain, and community partitioning can help group the enterprises by their degree of similarity and association in the supply chain network. Such partitioning helps to better understand the structure and organization of the supply chain, identify sets of businesses that have similar functions or roles, and thus provide a clearer view of subsequent analysis and decision making. For example, community partitioning results may show core enterprise groups in the supply chain and their associations, and may also identify specific types of enterprise collections, such as manufacturers, distributors, etc., that are important for developing targeted financing schemes and risk management strategies. In addition, the community division results may also provide network location information for each enterprise in the supply chain. Through community division, communities to which each enterprise belongs and the positions of the communities can be known, so that the status and the roles of the enterprises in the whole supply chain can be better understood. The information has important significance for evaluating the influence of the supply chain of enterprises, the exposure degree of risks and the relationship with other enterprises, and can provide more comprehensive and accurate data support for subsequent risk evaluation and financing scheme formulation.

Step S530, carrying out relation strength calculation according to the association information representation and the community division result, and obtaining the relation strength between enterprises by calculating the similarity in the association information;

In this step, the relationship strength between the enterprises is evaluated by an appropriate similarity calculation method using the attribute information, transaction data, and other related information between them. For example, cosine similarity or euclidean distance and other methods can be used for comparing attribute features between enterprises to evaluate the similarity degree between the enterprises; or the tightness of the trade relationship between enterprises may be evaluated according to the frequency and amount of the trade data. Through the calculation of the relationship strength, the relationship among enterprises, including cooperative relationship, competitive relationship and the like, can be more accurately understood. This is important for subsequent risk assessment and financing program formulation, and can help better identify other businesses that are closely related to the business, thereby better grasping the risk and opportunities in the supply chain. In addition, by considering community division results, the relationship strength between enterprises in different communities can be evaluated. This helps to understand more deeply the links and effects between different communities, providing more comprehensive and accurate data support for targeted formulation of financing schemes and risk management strategies.

And step S540, performing data visualization processing according to the network position and the relation strength to obtain a comprehensive data view.

The purpose of this step is to present the complex data information in an intuitive, easily understood manner, providing the user with a comprehensive, clear view of the supply chain network. In particular, the network location information of an enterprise in a supply chain network is utilized to determine its node location in the graph. This may be achieved by plotting nodes in two or three dimensions and grouping or coloring according to their network locations. For example, the core enterprise and the secondary enterprise may be marked as different node colors or shapes, respectively, to highlight their status in the supply chain. Such a visualization process can intuitively reveal the location and position of the enterprise in the supply chain, helping the user to quickly understand the overall structure of the supply chain. And secondly, determining the link weight or the thickness degree of the edges between the nodes by using the relationship strength between enterprises obtained through previous calculation. The relation strength among enterprises is reflected by adjusting the thickness or color and other attributes of the connection lines, so that the connection lines among nodes with stronger relation strength are more prominent, and the connection lines among nodes with weaker relation strength are thinner and weaker. Such a visualization process can intuitively exhibit the affinity of relationships between enterprises, helping users quickly understand the relationships and impact between various enterprises in the supply chain. Through this visualization process of the integrated data view, the user can clearly understand the structure of the supply chain network, the relationships between enterprises, and their status and influence throughout the supply chain. This helps the user to better understand the mechanism by which the supply chain operates, identify potential risks and opportunities, and provide more accurate and comprehensive data support for subsequent decisions.

The step S600 includes step S610, step S620, step S630, and step S640.

Step S610, modeling the time sequence data in the comprehensive data view based on a preset long-short-term memory network mathematical model, and obtaining a dynamic credit scoring time sequence of an enterprise by capturing long-term and short-term dependency relations in the enterprise credit scoring data;

It can be appreciated that the long-term memory network is a recurrent neural network, and is particularly suitable for processing and predicting time series data, and can effectively capture long-term dependency in time series. In this step, the business' credit score data is modeled using a long-term memory network to identify long-term trends and short-term fluctuations in the credit score time series. The correlation between time dependence and sequences in time sequence data can be considered through long-short term memory network modeling, so that future credit score trend can be predicted more accurately. The long-term and short-term memory network can automatically learn the mode and rule in the time series data and forecast the future development trend according to the historical data, so that the dynamic change of the credit score of the enterprise can be captured better.

Step S620, performing hidden state estimation on the dynamic credit scoring time sequence, and predicting the future credit change trend of the enterprise through randomness and trending in the state space model processing time sequence to obtain the future credit assessment result of the enterprise;

It should be noted that, hidden state estimation is a technique for deducing hidden states based on observation sequences, and is generally used in combination with a state space model. The state space model is a probabilistic model describing the evolution of the system, which divides the state of the system into an observed state and a hidden state, which is usually not directly observable, but rather is indirectly inferred from the observed state. In this step, the dynamic credit score time sequence of the enterprise is used as an observation sequence, and the hidden state is deduced through a hidden state estimation technology, namely, the state reflecting the potential factors and rules behind the credit score. The evolution law between these hidden states is then described using a state space model, taking into account randomness and trending in the time series. The dynamic change mechanism behind the credit score can be better understood through the state space model, so that the future credit change trend of the enterprise can be predicted more accurately. Finally, the future credit change trend of the enterprise is predicted, so that the future credit evaluation result of the enterprise can be obtained. This assessment may provide important reference information to financial institutions and businesses that help them better make risk management policies and financing decisions. By predicting the credit change trend of the enterprise in time, the financial institution can be helped to identify and manage the potential credit risk more effectively, and better financing opportunities and conditions are provided for the enterprise.

Step 630, performing bayesian inference processing by using a bayesian dynamic linear model according to the comprehensive data view and the credit evaluation result to obtain a risk prediction result, wherein the risk prediction result is the risk level of the enterprise at a specific future time point;

it can be understood that the bayesian dynamic linear model is a dynamic model based on bayesian statistical theory, and can continuously update the estimation of the unknown parameters by using the observation data under the condition of considering the prior knowledge. In the step, the comprehensive data view and the historical credit assessment result of the enterprise are used as observation data, and the Bayesian dynamic linear model is used for deducing the future risk level of the enterprise. Specifically, the comprehensive data view and the credit evaluation result of the enterprise are used as the input of a model, and the future risk level is predicted by using a Bayesian inference method according to the parameters of the model and combining the historical data and the current credit evaluation result. Such risk prediction results may provide important reference information for financial institutions and enterprises, helping them to better formulate risk management policies and financing decisions. By means of the inference processing of the Bayesian dynamic linear model, the risk level of the enterprise at a specific future time point can be accurately predicted by considering the influence of uncertainty factors and historical data. This helps financial institutions and enterprises discover and deal with potential risks in time, guaranteeing steady operation of financial systems and sustainable development of enterprises.

And step 640, analyzing connectivity and infectivity of the supply chain network by using a system risk model according to the comprehensive data view and the risk prediction result, and obtaining an evaluation systematic risk score by evaluating the influence of the enterprise on the whole network when the enterprise risks in the supply chain.

It will be appreciated that the system risk model is a model that comprehensively considers supply chain network structure, enterprise relevance, and risk propagation mechanisms, and can help understand the risk propagation process and the formation mechanism of systematic risks in supply chain systems. In this step, connectivity and infectivity of the supply chain network are analyzed using a system risk model to assess the extent of impact of the enterprise on the entire network when risks occur in the supply chain. Specifically, a specific risk event and impact range is first determined using the integrated data view and risk prediction results, and then a system risk model is used to simulate the propagation of risk in the supply chain network. By considering the relevance between enterprises, the structure of the supply chain network, and the path of risk propagation, the extent of impact on the overall supply chain system when an enterprise is at risk can be assessed. Finally, the influence degree of the enterprise on the whole network when the enterprise risks in the supply chain is obtained according to the evaluation result, and the influence degree is converted into a systematic risk score. This scoring may provide important reference information to financial institutions and businesses that help them better assess and manage systematic risks in the supply chain system, thereby reducing risk exposure and ensuring stable operation of the supply chain system.

The step S700 includes step S710, step S720, and step S730.

Step S710, carrying out modeling processing aiming at optimization targets of financing cost, risk minimization and liquidity maximization according to an evaluation result and a preset pareto front optimization mathematical model to obtain at least two preliminary financing schemes;

It will be appreciated that the pareto front optimization mathematical model is a multi-objective optimization method that can help find the best point of balance among a number of contradictory objectives, i.e. to maximize one objective without sacrificing the other. In this step, financing cost, risk and liquidity are regarded as three key targets in enterprise financing decisions, and they are uniformly modeled and processed by using a pareto front optimization mathematical model. Firstly, the weight and importance degree of financing cost, risk and liquidity are determined according to the evaluation result, and then, the pareto front optimization mathematical model is utilized to reasonably balance and optimize the targets. Through mathematical modeling and optimization processing, at least two preliminary financing schemes can be obtained, and the schemes can meet a plurality of targets and optimize the financing benefit and risk management of enterprises as much as possible. These preliminary financing schemes will provide a variety of options for the enterprise, helping it to formulate the most appropriate financing strategy according to the actual situation and needs. Meanwhile, by utilizing the pareto front edge optimization method, the mutual influence and the trade-off relation among different targets can be fully considered, so that the scientificity and the effectiveness of financing decisions are better realized.

Step S720, performing simulation processing according to all financing schemes, and obtaining an optimal financing scheme set by simulating different market environments and risk scenes by using a genetic algorithm and evaluating the performance of each preliminary financing scheme in various possible scenes;

Specifically, as shown in FIG. 2, all preliminary financing schemes are first simulated and placed as individuals into a population of genetic algorithms. Then, individuals in the population are optimized and evolved from generation to generation by simulating operations such as natural selection, crossover, mutation and the like until an optimal solution meeting specific conditions is found. In the simulation process, various possibilities under different market environments and risk scenarios are considered to evaluate the performance of each preliminary financing program under these scenarios. Through iterative optimization of a genetic algorithm, an optimal financing scheme set can be obtained, and the schemes have good performance under various situations and can meet the financing requirements and targets of enterprises to the greatest extent.

And step 730, processing the optimal financing scheme set based on preset deep learning, and performing strategy selection and dynamic adjustment processing by setting a reward mechanism and a strategy gradient method to obtain a final personalized financing scheme.

It should be noted that deep learning is an artificial intelligence technology, and can automatically discover rules and features in data by learning a large amount of data, and make effective decisions and predictions. In the step, the optimal financing scheme set is processed by using a preset deep learning model so as to further improve the quality and individuation degree of the financing scheme. Specifically, the optimal financing scheme set is first used as input data for the model, and then these data are learned and analyzed using a deep learning model to find potential rules and features therein. In the learning process, a corresponding rewarding mechanism is set to guide the model to evaluate and select different schemes, and meanwhile, a strategy gradient method is adopted to dynamically adjust and optimize the model. By constantly iterating learning and adjusting, the deep learning model can gradually improve understanding and optimizing capabilities of the financing scheme, so that the financing scheme which is more personalized and adapts to enterprise requirements is generated. The personalized schemes can better meet the actual conditions and targets of enterprises, and help the enterprises to achieve maximization of financing benefits and minimization of risks.

Example 2:

the embodiment provides a financing scheme generation system based on supply chain data, which comprises the following steps:

the identification module is used for carrying out identification processing according to the supply chain fusion data set, identifying key node enterprises in the supply chain through social network analysis and evaluating the centrality and connectivity of the key node enterprises in the whole supply chain to obtain a supply chain network model;

the integration module is used for carrying out data integration and association mining processing on the enterprise basic data, the supply chain influence scoring result and the supply chain network model based on a preset graph neural network mathematical model to obtain a comprehensive data view, wherein the comprehensive data view comprises the network position and the relationship strength of an enterprise applying for financial services in a supply chain;

the optimizing module is used for carrying out multi-objective optimizing processing on the evaluation result based on a preset deep learning mathematical model, generating at least two financing schemes under different scenes, and selecting and adjusting the financing schemes through a hybrid intelligent algorithm to obtain a final personalized financing scheme.

In the present application, the evaluation module includes:

The first evaluation unit is used for carrying out importance evaluation processing according to enterprise basic data and supply chain transaction data, and obtaining the importance score of an enterprise applying for financial services in a supply chain by constructing a transaction network diagram and calculating a PageRank value of each enterprise by adopting a PageRank algorithm;

The first analysis unit is used for carrying out redundant network analysis according to enterprise basic data and logistics information, and obtaining an alternative score by calculating the redundancy degree of each node enterprise in the supply chain and evaluating the position and influence of the application financial service enterprise in the logistics path;

The first processing unit is used for carrying out weighted comprehensive processing according to the importance scores and the substitutability scores to obtain a preliminary scoring result;

The first adjusting unit is used for adjusting according to the preliminary scoring result and the enterprise historical credit records in the financial institution data, and obtaining a final supply chain influence scoring result by calibrating the preliminary scoring result by using a weighted linear regression model.

In the present application, the fusion module includes:

the first clustering unit is used for carrying out clustering processing according to the supply chain transaction data and the logistics information, processing transaction amount, transaction frequency and logistics node information in the supply chain by utilizing a Gaussian mixture model, and fitting the multidimensional normal distribution of the data to obtain initial multidimensional data distribution;

the first calculation unit is used for carrying out condition summarization calculation according to the initial multidimensional data distribution and the financial institution data, classifying the loan interest rate, the credit approval standard and the repayment record by using a naive Bayesian classifier, and calculating the posterior probability of each category to obtain posterior probability sets of all categories;

the second processing unit is used for carrying out maximum posterior estimation processing according to the posterior probability set, and obtaining a fusion causal network reflecting the relevance among supply chain enterprises by constructing a causal relationship graph by utilizing a Bayesian network;

The first sampling unit is used for sampling and deducing the fusion causal network based on a preset Markov chain Monte Carlo mathematical model to obtain a supply chain fusion data set.

The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

The foregoing is merely illustrative of the present invention, and the present invention is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present invention. Therefore, the protection scope of the invention is subject to the protection scope of the claims.

Claims

1. A financing scheme generation method based on supply chain data, comprising:

performing multi-objective optimization processing on the evaluation result based on a preset deep learning mathematical model to generate at least two financing schemes under different scenes, and selecting and adjusting the financing schemes through a hybrid intelligent algorithm to obtain a final personalized financing scheme;

The data fusion processing is carried out on the second information by using a multi-source data fusion mathematical model based on Bayesian reasoning to obtain a supply chain fusion data set, which comprises the following steps:

Clustering is carried out according to the supply chain transaction data and the logistics information, transaction amount, transaction frequency and logistics node information in a supply chain are processed by using a Gaussian mixture model, and multidimensional normal distribution of the data is fitted to obtain initial multidimensional data distribution;

According to the initial multidimensional data distribution and the financial institution data, carrying out condition summarization calculation, classifying loan interest rates, credit approval standards and repayment records by using a naive Bayesian classifier, and calculating posterior probabilities of all the categories to obtain posterior probability sets of all the categories;

Performing maximum posterior estimation processing according to the posterior probability set, and constructing a causal relationship graph by using a Bayesian network to obtain a fusion causal network reflecting the relevance between supply chain enterprises;

sampling and deducing the fusion causal network based on a preset Markov chain Monte Carlo mathematical model to obtain a supply chain fusion data set;

the method comprises the steps of carrying out multi-objective optimization processing on the evaluation result based on a preset deep learning mathematical model, generating financing schemes under at least two different scenes, selecting and adjusting the financing schemes through a hybrid intelligent algorithm, and obtaining a final personalized financing scheme, wherein the method comprises the following steps:

According to the evaluation result and a preset pareto front optimization mathematical model, modeling processing is carried out aiming at optimization targets of financing cost, risk minimization and liquidity maximization to obtain at least two preliminary financing schemes;

Performing simulation processing according to all the financing schemes, and obtaining an optimal financing scheme set by simulating different market environments and risk scenes by using a genetic algorithm and evaluating the performance of each preliminary financing scheme in various possible scenes;

And processing the optimal financing scheme set based on preset deep learning, and performing strategy selection and dynamic adjustment processing by setting a reward mechanism and a strategy gradient method to obtain a final personalized financing scheme.

2. The supply chain data-based financing scheme generation method according to claim 1, wherein the evaluation processing is performed according to the first information and the second information, and the supply chain influence scoring result of the enterprise is obtained by evaluating the importance degree and the replaceability of the enterprise in the supply chain, comprising:

Carrying out importance evaluation processing according to the enterprise basic data and the supply chain transaction data, and obtaining importance scores of enterprises applying for financial services in a supply chain by constructing a transaction network diagram and calculating PageRank values of each enterprise by adopting a PageRank algorithm;

performing redundant network analysis according to the enterprise basic data and the logistics information, and obtaining an alternative score by calculating the redundancy degree of each node enterprise in a supply chain and evaluating the position and influence of an application financial service enterprise in a logistics path;

performing weighted comprehensive treatment according to the importance scores and the substitutability scores to obtain preliminary scoring results;

And adjusting according to the preliminary scoring result and the enterprise historical credit record in the financial institution data, and calibrating the preliminary scoring result by using a weighted linear regression model to obtain a final supply chain influence scoring result.

3. The supply chain data-based financing scheme generation method according to claim 1, wherein the identification processing is performed according to the supply chain fusion data set, key node enterprises in a supply chain are identified through social network analysis, centrality and connectivity of the key node enterprises in the whole supply chain are evaluated, and a supply chain network model is obtained, comprising:

Performing K core decomposition processing according to the supply chain fusion data set, and obtaining a preliminary key node set by removing low-level nodes;

Calculating the times of each node serving as a supply chain intermediate node according to the preliminary key node set to obtain a betweenness centrality score;

Calculating the degree and distance of each node according to the supply chain fusion data set and the betweenness centrality score to obtain a centrality score and a near centrality score of the node in a supply chain;

and carrying out weighted clustering coefficient calculation according to the betweenness centrality score, the degree centrality score and the approaching centrality score, obtaining the final centrality and connectivity of the key node enterprise by calculating the local network compactness of the node, and integrating the supply chain fusion data set to obtain a supply chain network model.

4. The supply chain data-based financing scheme generation method according to claim 1, wherein the data integration and association mining processing are performed on the enterprise base material, the supply chain influence scoring result and the supply chain network model based on a preset graph neural network mathematical model to obtain a comprehensive data view, and the method comprises the following steps:

Performing data integration and association mining processing on the enterprise basic data, the supply chain influence scoring result and the supply chain network model based on a preset graphic neural network mathematical model to obtain an association information representation;

Carrying out community division processing according to the associated information representation, and dividing enterprises closely associated with each other in the graph into the same community to obtain a community division result, wherein the community division result comprises network positions of the enterprises in a supply chain;

Carrying out relationship strength calculation according to the association information representation and the community division result, and obtaining relationship strength among enterprises by calculating similarity in the association information;

and carrying out data visualization processing according to the network position and the relation strength to obtain a comprehensive data view.

5. The supply chain data-based financing scheme generation method according to claim 1, wherein performing a time series analysis process according to the integrated data view to obtain an evaluation result comprises:

Modeling the time sequence data in the comprehensive data view based on a preset long-short-term memory network mathematical model, and obtaining a dynamic credit scoring time sequence of an enterprise by capturing long-term and short-term dependency relations in the enterprise credit scoring data;

Performing hidden state estimation on the dynamic credit scoring time sequence, and predicting future credit change trend of the enterprise through randomness and trend in the state space model processing time sequence to obtain future credit assessment results of the enterprise;

According to the comprehensive data view and the credit evaluation result, performing Bayesian inference processing by using a Bayesian dynamic linear model to obtain a risk prediction result, wherein the risk prediction result is the risk level of an enterprise at a specific future time point;

and analyzing connectivity and infectivity of the supply chain network by using a system risk model according to the comprehensive data view and the risk prediction result, and obtaining an evaluation systematic risk score by evaluating the influence of enterprises on the whole network when risks occur in the supply chain.

6. A financing scheme generation system based on supply chain data, comprising:

The optimizing module is used for carrying out multi-objective optimizing processing on the evaluation result based on a preset deep learning mathematical model, generating at least two financing schemes under different scenes, and selecting and adjusting the financing schemes through a hybrid intelligent algorithm to obtain a final personalized financing scheme;

wherein, the fusion module includes:

The first clustering unit is used for carrying out clustering processing according to the supply chain transaction data and the logistics information, processing transaction amount, transaction frequency and logistics node information in a supply chain by utilizing a Gaussian mixture model, and fitting the multidimensional normal distribution of the data to obtain initial multidimensional data distribution;

The first calculation unit is used for carrying out condition summarization calculation according to the initial multidimensional data distribution and the financial institution data, classifying loan interest rates, credit approval standards and repayment records by using a naive Bayesian classifier, and calculating posterior probabilities of all categories to obtain posterior probability sets of all categories;

The first sampling unit is used for sampling and deducing the fusion causal network based on a preset Markov chain Monte Carlo mathematical model to obtain a supply chain fusion data set;

wherein, the optimization module includes:

The first modeling unit is used for performing modeling processing on optimization targets of financing cost, risk minimization and liquidity maximization according to the evaluation result and a preset pareto front optimization mathematical model to obtain at least two preliminary financing schemes;

The first simulation unit is used for performing simulation processing according to all the financing schemes, and obtaining an optimal financing scheme set by simulating different market environments and risk scenes by using a genetic algorithm and evaluating the performance of each preliminary financing scheme under various possible scenes;

the first generation unit is used for processing the optimal financing scheme set based on preset deep learning, and performing strategy selection and dynamic adjustment processing by setting a reward mechanism and a strategy gradient method to obtain a final personalized financing scheme.

7. The supply chain data-based financing scheme generation system of claim 6, wherein the evaluation module comprises:

the first evaluation unit is used for carrying out importance evaluation processing according to the enterprise basic data and the supply chain transaction data, and obtaining importance scores of the enterprises applying for the financial services in the supply chain by constructing a transaction network diagram and calculating PageRank values of each enterprise by adopting a PageRank algorithm;

the first analysis unit is used for carrying out redundant network analysis according to the enterprise basic data and the logistics information, and obtaining an alternative score by calculating the redundancy degree of each node enterprise in a supply chain and evaluating the position and influence of an application financial service enterprise in a logistics path;

And the first adjusting unit is used for adjusting according to the preliminary scoring result and the enterprise historical credit records in the financial institution data, and obtaining a final supply chain influence scoring result by calibrating the preliminary scoring result by using a weighted linear regression model.