Financial product intelligent matching system based on multi-source data integration

Abstract

The invention discloses a financial product intelligent matching system based on multi-source data integration, and relates to the technical field of finance, and the system comprises a data access module which is used for collecting enterprise heterogeneous data from an industrial end and a multi-source system; the data cleaning and structuring module is used for cleaning, integrating and normalizing the heterogeneous data and generating enterprise label data according to a data system; the rule configuration module is used for configuring and maintaining an access rule based on the financial product and converting the access rule into a computable label; the intelligent matching module is used for realizing bidirectional matching by calculating the similarity between an enterprise and a product based on an enterprise label and a product access rule, and the business opportunity management and collaboration module is used for carrying out business opportunity scoring and level-to-level management on a matching result. According to the application, various data systems such as ERP, tax affairs, credit investigation, supply chains and the like are seamlessly connected, and integration and real-time acquisition of heterogeneous data such as enterprise operation data, credit rating, cash flow, asset information and the like at an industrial end are realized, so that data sources are more comprehensive and timely.

Description

Technical Field

[0001] The present invention relates to the field of financial technology, and in particular to an intelligent matching system for financial products based on multi-source data integration. Background Art

[0002] In the financial sector, user needs are becoming increasingly diverse, and the traditional "product launch-user matching" model is no longer able to meet the demand for personalized services. Intelligent financial product matching systems leverage technologies such as big data, machine learning, and artificial intelligence to build user profiles based on multi-dimensional data such as basic user information, investment preferences, and risk tolerance. These profiles then integrate the product's return, risk, maturity, and market environment to provide precise product recommendations. This not only enhances the user experience but also provides financial institutions with more effective marketing tools, reducing promotion costs and increasing product conversion rates.

[0003] Publication No. CN110097430A discloses an AI-based intelligent matching system for auto finance products. The system includes a matching model generation module for individual auto finance products and a matching model generation module for the optimal auto finance product for new users. The matching model generation module for individual auto finance products includes a first information acquisition unit, which is connected to a data processing unit via a signal, which is connected to a matching generation unit via a signal, which is connected to a comparison unit via a signal. This AI-based intelligent matching system for auto finance products offers the advantages of intelligent and rapid screening, reducing staff workload and improving the efficiency of auto finance product selection. Large-scale screening reduces screening errors, enables the selection of optimal auto finance products, and improves the quality rate of auto finance products.

[0004] During the financing process, significant information asymmetry often exists between businesses and financial institutions. Information such as a business's operating conditions, credit rating, and cash flow status is difficult to convey to financial institutions in a timely and accurate manner. Consequently, credit decisions often rely on crude credit standards or extensive manual review. This often leads to delays in funding due to cumbersome screening processes and lengthy review cycles. Furthermore, due to significant differences in industry, scale, operating model, and growth stage among businesses, a single financial product or standardized credit assessment model is unlikely to address all needs. Summary of the Invention

[0005] One of the purposes of the present invention is to provide an intelligent matching system for financial products based on multi-source data integration, which is aimed at financial institutions and industrial enterprises, and builds an intelligent matching platform for enterprise entity data and financial product access rules. Through standardized enterprise data tags and dynamic product rule engines, it can achieve efficient matching of enterprise needs and financial products, accumulate business opportunity clues and improve the accuracy of asset allocation of financial institutions.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a financial product intelligent matching system based on multi-source data integration, comprising: The data access module is used to collect heterogeneous enterprise data from the industry end and multi-source systems: industrial and commercial information, operating data, credit ratings, and asset information; The data cleaning and structuring module cleans, integrates, and normalizes heterogeneous data, and generates enterprise label data and enterprise portraits based on the data system; The rule configuration module configures and maintains access rules based on financial products and converts them into computable tags; The intelligent matching module achieves two-way matching by calculating the similarity between enterprises and products based on enterprise tags and product access rules. Forward matching screens eligible financial products for enterprises, while reverse matching identifies target enterprises for financial products. The business opportunity management and collaboration module scores and manages matching results in different levels, implements follow-up and feedback on business opportunities through automated allocation and closed-loop tracking, and dynamically optimizes the matching algorithm based on feedback data.

[0007] In one or more embodiments of the present invention, a multi-source system is connected to the industry-side ERP, tax, credit reporting and supply chain systems to collect corporate business information, operating data, credit ratings and core fields of asset collateral. The data access module is configured with multiple data connection methods, including API interface, file transfer and direct database connection.

[0008] In one or more embodiments of the present invention, the data cleaning and structuring module classifies, summarizes and standardizes enterprise data by defining a unified enterprise label standard system, and generates structured labels including industry classification, enterprise size, financing purpose and risk level.

[0009] In one or more embodiments of the present invention, a unified enterprise labeling standard system is defined and an enterprise model is established, which includes the following dimensions based on the different attributes and business needs of enterprises: Industry classification, based on the industry in which the enterprise is located and combined with industry standards; Enterprise size, which is divided into size levels based on the company's annual revenue, number of employees, and total assets; Financing purpose: determine the enterprise's financing needs and classify them; Risk level, build a risk scoring model based on the company's credit rating, financial health, and operational stability.

[0010] In one or more embodiments of the present invention, the steps for constructing an enterprise risk scoring module are as follows: Obtain cleaned corporate credit rating, financial health, and operational stability data; Construct key indicators that affect enterprise risks to form a feature vector: Numericalization of credit ratings, converting credit ratings into numerical indicators; Refine financial indicators, calculate comprehensive indicators such as debt-to-asset ratio, operating profit margin, and cash flow ratio, and introduce dynamic indicators such as volatility and year-on-year growth rate to reflect the health of the company's operations; Constructing operational stability characteristics, extracting the annualized growth rate and volatility coefficient of operating income and net profit over the years, and establishing operational stability indicators; Determine the relationship between each characteristic and risk through correlation analysis; A risk scoring model is established based on the feature vector to determine the enterprise risk, a risk label is constructed for the enterprise based on historical risk events, and the enterprise is divided into different risk levels based on the risk score.

[0011] In one or more embodiments of the present invention, the difference between the enterprise industry label and the actual operating status is determined based on the industry enterprise scale and enterprise revenue status: Collect data on the size of companies in the industry and accounts receivable data; Use historical data and industry statistics to determine the range of accounts receivable data that companies of all sizes should have under normal operating conditions; Compare the target enterprise's scale indicators and accounts receivable status, quantitatively compare the target enterprise's accounts receivable status with the benchmark status of enterprises of the same size, use the accounts receivable to operating income ratio and accounts receivable turnover rate indicators to calculate the degree of deviation from the industry range, set thresholds based on the degree of deviation, and mark enterprises that exceed the set thresholds as abnormal operations.

[0012] In one or more embodiments of the present invention, the similarity between enterprises and products is calculated to achieve two-way matching: The similarity between the enterprise label vector and the product rule vector is measured by cosine similarity, and the indicator weight is determined based on statistical analysis. Forward matching: Calculate the similarity between the enterprise label vector and the access rule vector of each financial product, screen financial products based on the similarity score, sort the candidate products according to the matching score, and recommend companies with high matching scores first; Reverse matching: Build a product vector based on product access rules, calculate the similarity between companies and the product, and mark companies within the similarity threshold as potential customers. By complementing each other through forward matching and reverse matching, and continuously optimizing the matching model through business feedback, the weights and matching thresholds of each dimension are adjusted.

[0013] In one or more embodiments of the present invention, the similarity between products and enterprises is calculated based on Yuxuan similarity, and the cosine value of the angle between two vectors is measured to calculate the cosine similarity. : ; in, Represents the dot product of the enterprise vector k and the product vector r: ; Represents the Euclidean norm of the enterprise vector k and product vector r: .

[0014] In one or more embodiments of the present invention, a target enterprise obtains a target product recommendation through forward matching, and a target product identifies a target enterprise in reverse matching. The matching results on both ends are cross-compared, and when the matching results show a high similarity in both the forward and reverse directions, the matching results are marked; Among them, forward matching recommends target products, and in reverse matching, when the similarity between the target product and the target enterprise is outside the matching threshold, data deviation or abnormal matching rules are marked.

[0015] In one or more embodiments of the present invention, a weighted comprehensive business opportunity score is calculated, and business opportunities are divided into different levels based on the calculated total score. Based on the geographical location of each branch of the financial institution, the professional tags of the relationship manager, and historical performance, high-quality business opportunities are automatically assigned to the most suitable relationship manager, thereby establishing a feedback closed-loop system: The dashboard displays the deviation between opportunity scores and actual conversions in real time; Dynamically adjust the parameters of the matching algorithm based on the results of statistical analysis; When updating an algorithm, first verify the effectiveness of the new algorithm on a small scale before promoting it to the entire system.

[0016] Through the above technical solution, the present invention has the following beneficial effects: 1. This application seamlessly connects with multi-source data systems such as ERP, taxation, credit reporting, and supply chain, and realizes the integration and real-time collection of heterogeneous data such as business operating data, credit ratings, cash flow, and asset information of industrial enterprises, making the data source more comprehensive and timely, and standardizing the scattered information according to a unified dimension, thereby forming an accurate corporate portrait and providing a more scientific basis for matching.

[0017] 2. The use of a dynamic rule engine allows financial institutions to adjust rule weights based on real-time market conditions, risk preferences, and product characteristics, making the matching process more flexible and adaptable. It not only achieves positive matching, but also introduces reverse matching logic, that is, actively scanning the enterprise database for high-value or specialized financial products and looking for potential high-quality customers.

[0018] 3. Generate matching results, manage business opportunities in a hierarchical manner, and push high-quality business opportunities to suitable account managers or financial institution departments through an automated allocation mechanism, thereby improving business execution efficiency and resource utilization, and solving the problems of scattered business opportunity management and inefficient follow-up in the existing system.

[0019] 4. Conduct risk assessments at an early stage of matching enterprises with financial products, automatically eliminate enterprises that do not meet risk qualifications, implement pre-emptive risk management, reduce non-performing loan risks, and establish a comprehensive enterprise portrait and multi-dimensional labeling system. The system can tailor matching plans based on the specific circumstances of the enterprise, making financial product recommendations more personalized.

[0020] Other features and advantages of the present invention will be described in the following description, and part of them will become obvious from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Schematic diagram of the system of the present invention. DETAILED DESCRIPTION

[0022] The following drawings illustrate various embodiments of the present invention. For clarity, many practical details are included in the following description. However, it should be understood that these practical details are not intended to limit the present invention. In other words, in some embodiments of the present invention, these practical details are optional. Furthermore, features from different embodiments may be interchangeably applicable, where practically possible.

[0023] Unless otherwise defined, all terms used herein (including technical and scientific terms) have their ordinary meanings as understood by those skilled in the art. Furthermore, the definitions of the aforementioned terms in commonly used dictionaries should be interpreted in the context of this specification as consistent with the meanings in the art relevant to the present invention. Unless otherwise explicitly defined, these terms should not be interpreted as having idealized or overly formal meanings.

[0024] See Figure 1 As shown, the present invention provides a financial product intelligent matching system based on multi-source data integration, which realizes two-way matching logic based on multi-source data integration and the access rules of financial institutions for configuring products.

[0025] The financial product intelligent matching system includes: The data access module is used to collect heterogeneous enterprise data from the industry end and multi-source systems: industrial and commercial information, operating data, credit ratings, and asset information; The data cleaning and structuring module cleans, integrates, and normalizes heterogeneous data, and generates enterprise label data and enterprise portraits based on the data system; The rule configuration module configures and maintains access rules based on financial products and converts them into computable tags; The intelligent matching module achieves two-way matching by calculating the similarity between enterprises and products based on enterprise tags and product access rules. Forward matching screens eligible financial products for enterprises, while reverse matching identifies target enterprises for financial products. The business opportunity management and collaboration module scores and manages matching results in different levels, implements follow-up and feedback on business opportunities through automated allocation and closed-loop tracking, and dynamically optimizes the matching algorithm based on feedback data.

[0026] In one feasible approach, a matching algorithm is used to reduce the asymmetry of corporate financing information so that the matching of financial resources can be quickly carried out. On the basis of achieving forward matching, reverse matching logic is introduced to actively scan the corporate database for high-value or special financial products and find potential high-quality customers. By determining the similarity, high-precision matching between enterprises and financial products can be achieved.

[0027] It adopts the intelligence and automation of multiple links including standardization, matching logic, risk control, business opportunity accumulation and closed-loop feedback. With the help of multi-source data integration, dynamic rule engine, two-way matching strategy and continuous optimization mechanism, it greatly improves the matching efficiency and accuracy, and builds a new ecosystem of risk control and personalized services for enterprises and financial institutions.

[0028] In one embodiment, a multi-source system connects to the industry-side ERP, tax, credit reporting, and supply chain systems to collect core fields of corporate business information, operating data, credit ratings, and asset collateral. The data access module is configured with multiple data connection methods, including API interfaces, file transfers, and direct database connections.

[0029] In one feasible approach, by connecting multiple data systems such as ERP, taxation, credit reporting, and supply chain through multi-source systems, the integration and real-time collection of heterogeneous data such as industrial enterprise operating data, credit ratings, cash flow, and asset information can be achieved, making the data source more comprehensive and timely.

[0030] This avoids the need for financial institutions to conduct corresponding data queries on enterprises when establishing financing relationships with enterprises, improves the timeliness and comprehensiveness of enterprise data, and further makes it more convenient for financial institutions to conduct audits of enterprises.

[0031] In one embodiment, the data cleaning and structuring module classifies, summarizes and standardizes enterprise data by defining a unified enterprise label standard system, and generates structured labels including industry classification, enterprise size, financing purpose and risk level.

[0032] In one feasible approach, a unified labeling standard system eliminates barriers in heterogeneous data formats, allowing multi-source data to be efficiently integrated after cleaning, forming an authoritative and credible corporate portrait, and directly improving the accuracy and efficiency of the matching algorithm.

[0033] Among them, structured tags provide a refined basis for subsequent dynamic rule matching and business opportunity evaluation, enabling the system to quickly identify target enterprises that meet the requirements of financial products among a large number of enterprises.

[0034] In one embodiment, a unified enterprise labeling standard system is defined to establish an enterprise model, which includes the following dimensions based on the different attributes and business needs of the enterprise: Industry classification, based on the industry in which the enterprise is located and combined with industry standards; Enterprise size, which is divided into size levels based on the company's annual revenue, number of employees, and total assets; Financing purpose: determine the enterprise's financing needs and classify them; Risk level, build a risk scoring model based on the company's credit rating, financial health, and operational stability.

[0035] Exemplary: Industry classification, collects public information released by enterprises, including enterprise registration information, business scope description and industry reports, parses the text information, extracts core keywords, maps them to standard industry classifications, and designs a multi-label mechanism for labeling to ensure that each enterprise label standard reflects its main business attributes.

[0036] Enterprise scale: collect and summarize the financial statements, tax data and employee information of enterprises to ensure the diversity and accuracy of data sources, establish specific scale classification criteria, consider the number of employees and total assets, and adopt a comprehensive evaluation of multiple indicators to ensure full adaptability to the characteristics and business models of various industries.

[0037] Financing purpose: By collecting and analyzing multi-dimensional data such as the company's past financing cases, business plans, project application materials, and financial analysis reports, we establish predefined financing purpose categories, and use classification algorithms to classify the company's financing needs into corresponding categories. We dynamically adjust and optimize the classification standards to ensure that the classification results are consistent with actual business needs.

[0038] Risk level: collect data such as the company's credit report, audit report, financial statements, and industry risk indicators, quantify key indicators, calculate risk scores, set quantitative thresholds for risk levels, and regularly update the model through retrospective analysis and market adjustment feedback to ensure the real-time and accuracy of risk assessment.

[0039] In one embodiment, the steps for constructing an enterprise risk scoring module are as follows: Obtain cleaned corporate credit rating, financial health, and operational stability data; Construct key indicators that affect enterprise risks to form a feature vector: Numericalization of credit ratings, converting credit ratings into numerical indicators; Refine financial indicators, calculate comprehensive indicators such as debt-to-asset ratio, operating profit margin, and cash flow ratio, and introduce dynamic indicators such as volatility and year-on-year growth rate to reflect the health of the company's operations; Constructing operational stability characteristics, extracting the annualized growth rate and volatility coefficient of operating income and net profit over the years, and establishing operational stability indicators; Determine the relationship between each characteristic and risk through correlation analysis; A risk scoring model is established based on the feature vector to determine the enterprise risk, a risk label is constructed for the enterprise based on historical risk events, and the enterprise is divided into different risk levels based on the risk score.

[0040] In one feasible approach, a statistical model, a machine learning model, or a combined model may be selected based on data characteristics to establish a risk scoring model.

[0041] Combine historical data for model training, improve the accuracy of model predictions through optimization, measure model performance, and develop a specific risk scoring system based on model output.

[0042] The ROC curve, AUC value, accuracy, recall rate, and F1 value indicators are used to comprehensively evaluate the model, analyze the confusion matrix, understand the model's prediction situation in each risk category, normalize the probability or score output by the model to a certain scoring range, and set the breakpoints for risk level division based on expert opinions and historical data.

[0043] In one embodiment, the difference between the enterprise industry label and the actual operating status is determined based on the industry enterprise scale and enterprise revenue status: Collect data on enterprise size and accounts receivable within the industry; Use historical data and industry statistics to determine the range of accounts receivable data that companies of all sizes should have under normal operating conditions; Compare the target enterprise's scale indicators and accounts receivable status, quantitatively compare the target enterprise's accounts receivable status with the benchmark status of enterprises of the same size, use the accounts receivable to operating income ratio and accounts receivable turnover rate indicators to calculate the degree of deviation from the industry range, set thresholds based on the degree of deviation, and mark enterprises that exceed the set thresholds as abnormal operations.

[0044] In one feasible approach, by building an industry benchmark model and comparing the company size with the accounts receivable status, we can effectively identify companies whose accounts receivable management does not match the scale characteristics of the same industry. By using data mining and comparison mechanisms, we can provide quantitative indicators and judgment standards for determining whether there is a discrepancy between the industry labeling of a company and its actual operation, thereby avoiding abnormal operations of the company.

[0045] By comparing the scale characteristics and accounts receivable management levels of companies in the industry at a macro level, and combining the companies' own financial data, we can identify cases where the scale of the companies and their accounts receivable performance are "out of line", thereby providing data support for judging the actual business activities of the companies and avoiding the problem of companies operating in a superficial manner.

[0046] For example, large enterprises in the industry usually have a lower proportion of accounts receivable and a faster turnover due to their large business volume, while small enterprises may have a higher proportion of accounts receivable or a lower turnover rate due to reasons such as higher credit risk.

[0047] One of the large enterprises has operating income and total assets that meet the requirements of large enterprises in the industry, but its accounts receivable ratio is far higher than the industry standard and its turnover is significantly low, indicating that there are problems with its actual operating model or business transformation, loose credit management, etc.

[0048] For small or medium-sized enterprises, if the accounts receivable indicator is abnormally low, it indicates that their credit management is abnormally strict and their actual business model tends to be a business with fast collection speed, which deviates from the registered industry standards.

[0049] After identifying abnormally operating enterprises, further investigations will be conducted on the enterprises to reduce financial financing risks.

[0050] In one embodiment, the similarity between enterprises and products is calculated to achieve two-way matching: The similarity between the enterprise label vector and the product rule vector is measured by cosine similarity, and the indicator weight is determined based on statistical analysis. Forward matching: Calculate the similarity between the enterprise label vector and the access rule vector of each financial product, screen financial products based on the similarity score, sort the candidate products according to the matching score, and recommend companies with high matching scores first; Reverse matching: Build a product vector based on product access rules, calculate the similarity between companies and the product, and mark companies within the similarity threshold as potential customers. By complementing each other through forward matching and reverse matching, and continuously optimizing the matching model through business feedback, the weights and matching thresholds of each dimension are adjusted.

[0051] One feasible approach involves a two-way matching mechanism based on enterprise tags and product access rules, which accurately matches enterprises and products by calculating similarities. Forward matching enables enterprises to quickly receive recommendations for highly compatible financial products, while reverse matching provides financial institutions with the ability to proactively identify potential target enterprises, enabling information sharing, efficient resource allocation, and business optimization.

[0052] Bidirectional verification improves the accuracy and robustness of the overall matching by comparing the matching results in two directions, and provides data feedback and guidance for subsequent model optimization.

[0053] In one embodiment, the similarity between products and enterprises is calculated based on Yuxuan similarity, which measures the cosine value of the angle between two vectors and calculates the cosine similarity. : ; in, Represents the dot product of the enterprise vector k and the product vector r: ; Represents the Euclidean norm of the enterprise vector k and product vector r: .

[0054] In one implementation, cosine similarity measures how similar two vectors are in direction, regardless of their magnitude. When two vectors are in exactly the same direction, the cosine similarity is 1; when they are orthogonal (unrelated), the value is 0; and if they are in completely opposite directions, the result is -1.

[0055] For example, the label vector of an enterprise is k=[0.8, 0.6, 0.9, 0.7], which respectively represents the normalized scores of industry classification, enterprise size, financing purpose, and risk level. The vector obtained after quantification of the access rules of financial products is r=[0.9, 0.6, 0.8, 0.6].

[0056] Substituting the cosine similarity calculation formula into the result is 0.9946, which is very high (close to 1), indicating that the enterprise label and product rules are very consistent in all dimensions. A threshold (0.8 or 0.85) is set to determine whether the matching conditions are met.

[0057] In one embodiment, the target enterprise obtains target product recommendations through forward matching, and the target product identifies the target enterprise in reverse matching. The matching results on both ends are cross-compared. When the matching results show high similarity in both forward and reverse directions, the matching results are marked; Among them, the forward matching recommends the target product, but when the similarity between the target product and the target enterprise in the reverse matching is outside the matching threshold, the data deviation or abnormal matching rule is marked.

[0058] Among them, the two-way verification mechanism of forward matching and reverse matching can form a closed-loop process of cross-confirmation, which not only improves the matching efficiency, but also makes the matching system more data-intelligent and self-correcting, providing strong support for precise services in financial technology scenarios.

[0059] For example, assume that the standardized label vector of enterprise k is [0.8, 0.6, 0.9, 0.7] and the admission rule vector of product r is [0.9, 0.6, 0.8, 0.6].

[0060] Forward matching calculation: The matching score between enterprise k and product r (similarity 0.92) exceeds the threshold set by the system (0.9), then k is recommended this product.

[0061] Reverse matching calculation: Similarly, the system calculates the matching between product r and each enterprise among all enterprises, and finds that enterprise k is also in a high similarity segment (matching score 0.91), marking k as a high-quality target for the product.

[0062] Bidirectional Verification: When r receives a high match score in both the forward and reverse matching, the system confirms it as a truly compatible match. If there is a significant discrepancy between the two results, further investigation or adjustment of the matching rules will be prompted.

[0063] In one embodiment, a weighted comprehensive opportunity score is calculated and divided into different levels based on the total score. High-quality opportunities are automatically assigned to the most suitable account managers based on the geographical location of each financial institution's branches, the account managers' professional tags, and historical performance, thus establishing a closed-loop feedback system: The dashboard displays the deviation between opportunity scores and actual conversions in real time; Dynamically adjust the parameters of the matching algorithm based on the results of statistical analysis; When updating an algorithm, first verify the effectiveness of the new algorithm on a small scale before promoting it to the entire system.

[0064] In one feasible approach, through sophisticated business opportunity scoring and tiered management, an automated allocation mechanism, and a closed-loop tracking and feedback system, full-process management of matching results can be achieved. By using real-time feedback data, not only can the accuracy of forward and reverse matching be verified, but the matching algorithm can also be continuously iterated and optimized based on actual operating performance, thereby improving the success rate of financing matching and risk management capabilities.

[0065] Although the present invention is disclosed in conjunction with the above embodiments, it is not intended to limit the present invention. Anyone familiar with this technology can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention should be based on the definition of the attached claims.

Claims

1. A financial product intelligent matching system based on multi-source data integration, characterized by: include: The data access module is used to collect heterogeneous enterprise data from the industry end and multi-source systems: industrial and commercial information, operating data, credit ratings, and asset information; The data cleaning and structuring module cleans, integrates, and normalizes heterogeneous data, and generates enterprise label data and enterprise portraits based on the data system; The rule configuration module configures and maintains access rules based on financial products and converts them into computable tags; The intelligent matching module achieves two-way matching by calculating the similarity between enterprises and products based on enterprise tags and product access rules. Forward matching screens eligible financial products for enterprises, while reverse matching identifies target enterprises for financial products. The business opportunity management and collaboration module scores and manages matching results in different levels, implements follow-up and feedback on business opportunities through automated allocation and closed-loop tracking, and dynamically optimizes the matching algorithm based on feedback data.

2. The financial product intelligent matching system based on multi-source data integration according to claim 1, characterized in that: The multi-source system connects to the industry's ERP, tax, credit reporting and supply chain systems to collect corporate business information, operating data, credit ratings and core fields of asset collateral. The data access module is equipped with multiple data connection methods, including API interface, file transfer and direct database connection.

3. The financial product intelligent matching system based on multi-source data integration according to claim 2, characterized in that: The data cleaning and structuring module classifies, summarizes and standardizes enterprise data by defining a unified enterprise label standard system, generating structured labels including industry classification, enterprise size, financing purpose and risk level.

4. The financial product intelligent matching system based on multi-source data integration according to claim 3 is characterized in that: Define a unified enterprise labeling standard system and establish an enterprise model, including the following dimensions based on the different attributes and business needs of enterprises: Industry classification, based on the industry in which the enterprise is located and combined with industry standards; Enterprise size, which is divided into size levels based on the company's annual revenue, number of employees, and total assets; Financing purpose: determine the enterprise's financing needs and classify them; Risk level, build a risk scoring model based on the company's credit rating, financial health, and operational stability.

5. The financial product intelligent matching system based on multi-source data integration according to claim 4 is characterized in that: The steps to build the enterprise risk scoring module are as follows: Obtain cleaned corporate credit rating, financial health, and operational stability data; Construct key indicators that affect enterprise risks to form a feature vector: Numericalization of credit ratings, converting credit ratings into numerical indicators; Refine financial indicators, calculate comprehensive indicators such as debt-to-asset ratio, operating profit margin, and cash flow ratio, and introduce dynamic indicators such as volatility and year-on-year growth rate to reflect the health of the company's operations; Constructing operational stability characteristics, extracting the annualized growth rate and volatility coefficient of operating income and net profit over the years, and establishing operational stability indicators; Determine the relationship between each characteristic and risk through correlation analysis; A risk scoring model is established based on the feature vector to determine the enterprise risk, a risk label is constructed for the enterprise based on historical risk events, and the enterprise is divided into different risk levels based on the risk score.

6. The financial product intelligent matching system based on multi-source data integration according to claim 5, characterized in that: Based on the industry enterprise scale and enterprise revenue status, determine the difference between the enterprise industry label and actual operating status: Collect data on the size of companies in the industry and accounts receivable data; Use historical data and industry statistics to determine the range of accounts receivable data that companies of all sizes should have under normal operating conditions; Compare the target enterprise's scale indicators and accounts receivable status, quantitatively compare the target enterprise's accounts receivable status with the benchmark status of enterprises of the same size, use the accounts receivable to operating income ratio and accounts receivable turnover rate indicators to calculate the degree of deviation from the industry range, set thresholds based on the degree of deviation, and mark enterprises that exceed the set thresholds as abnormal operations.

7. The financial product intelligent matching system based on multi-source data integration according to claim 6, characterized in that: Calculate the similarity between enterprises and products to achieve two-way matching: The similarity between the enterprise label vector and the product rule vector is measured by cosine similarity, and the indicator weight is determined based on statistical analysis. Forward matching: Calculate the similarity between the enterprise label vector and the access rule vector of each financial product, screen financial products based on the similarity score, sort the candidate products according to the matching score, and recommend companies with high matching scores first; Reverse matching: Build a product vector based on product access rules, calculate the similarity between companies and the product, and mark companies within the similarity threshold as potential customers. By complementing each other through forward matching and reverse matching, and continuously optimizing the matching model through business feedback, the weights and matching thresholds of each dimension are adjusted.

8. The financial product intelligent matching system based on multi-source data integration according to claim 7, characterized in that: Based on Yuxuan similarity, the similarity between products and enterprises is calculated, which measures the cosine value of the angle between two vectors and calculates the cosine similarity. : ； in, Represents the dot product of the enterprise vector k and the product vector r: ； Represents the Euclidean norm of the enterprise vector k and product vector r: 。 9. The financial product intelligent matching system based on multi-source data integration according to claim 8, characterized in that: The target enterprise obtains the target product recommendation through forward matching, and the target product identifies the target enterprise in reverse matching. The matching results on both ends are cross-compared. When the matching results show a high similarity in both forward and reverse directions, the matching results are marked; Among them, forward matching recommends target products, and in reverse matching, when the similarity between the target product and the target enterprise is outside the matching threshold, data deviation or abnormal matching rules are marked.

10. The financial product intelligent matching system based on multi-source data integration according to claim 9, characterized in that: By calculating a weighted comprehensive opportunity score, we categorize opportunities into different levels based on the total score. Based on the geographical location of each financial institution's branches, the relationship managers' professional tags, and historical performance, we automatically assign high-quality opportunities to the most suitable relationship managers, thus building a closed-loop feedback system. The dashboard displays the deviation between opportunity scores and actual conversions in real time; Dynamically adjust the parameters of the matching algorithm based on the results of statistical analysis; When updating an algorithm, first verify the effectiveness of the new algorithm on a small scale before promoting it to the entire system.

Images