A financial risk control evaluation method and system based on multi-source data fusion

Abstract

The application discloses a kind of financial risk control evaluation method and system based on multi-source data fusion, comprising: access internal, third party and low-credibility supplementary data and clean up;With matching rate, completeness, freshness, fluctuation coefficient quantification accuracy, integrity, timeliness, stability, combined with type coefficient λ, correlation degree μ, sensitivity σ to calculate the score of credibility with subjective and objective coupling algorithm, according to layer distribution fusion weight and extract low-quality data risk signal;Support visual self-defined index and hierarchical model;Splice feature calculation score and map risk level, realize pre-event / middle / post-event dynamic update;Matching credit policy, when the credibility change ≥5 points or high priority risk, dynamically adjust weight.The application includes data governance, index modeling, full-cycle scoring, intelligent decision-making, traceability audit module, to improve evaluation accuracy and compliance.

Description

Technical Field

[0001] This invention belongs to the field of financial technology and big data processing technology, specifically relating to a financial risk control assessment method and system based on multi-source data fusion. Background Technology

[0002] With the acceleration of digital transformation in the financial sector, multi-source data fusion has become a core technological direction for credit risk assessment in financial institutions. Current mainstream risk control solutions generally integrate multi-dimensional data such as internal transaction records, credit records, third-party credit reports, business and tax information, and judicial and public opinion data. However, in practical applications, three unresolved technical pain points remain:

[0003] 1. Difficulty in integrating multi-source data: The credibility of internal transaction data, third-party credit data, and public opinion data varies significantly, there is a lack of unified quantitative evaluation standards, and weight adjustments lag far behind changes in the quality of data sources;

[0004] 2. Poor scenario adaptability: Existing solutions are either only for personal loans or only cover corporate credit, with fixed indicators and models, which cannot quickly adapt to the differentiated risk control needs of different customer groups such as SMEs and science and technology innovation enterprises;

[0005] 3. Lack of full-cycle monitoring: Most assessments remain at the pre-loan access stage, with insufficient ability to track risk triggers during the process and post-loan performance. Key risk signals in low-reliability data are easily ignored or excessively interfere with the scoring results.

[0006] While patent CN122022979A achieves credibility stratification and dynamic weight adjustment, it does not support user-defined indicators and models, resulting in weak scenario scalability. Patent CN122048505A, while supporting full-cycle evaluation and custom configuration, fails to address the issues of quantifying the credibility of multi-source data and extracting risk signals from low-quality data. Neither patent forms a closed-loop solution encompassing "data quality control - flexible model configuration - full-cycle dynamic decision-making." Summary of the Invention

[0007] The purpose of this invention is to provide a financial risk control assessment method and system based on multi-source data fusion to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a financial risk control assessment method based on multi-source data fusion, comprising the following steps:

[0009] Access to three types of data sources: internal data from financial institutions, third-party compliant data, and low-reliability supplementary data, and complete format unification and cleaning processes;

[0010] The accuracy, completeness, timeliness, and stability of each data source are quantified by data matching rate, field completeness rate, data freshness, and data quality fluctuation coefficient, and a comprehensive credibility score is calculated by combining multi-dimensional constraint parameters.

[0011] The data is divided into three layers—high, medium, and low—based on its overall credibility score, and fusion weights are assigned accordingly. Key risk signals are extracted from the low-credibility data.

[0012] Respond to user operations by customizing risk control indicators and calculation rules, and construct a hierarchical risk control model that includes multiple assessment modules;

[0013] The data features and risk signals of each layer are spliced ​​together, and the raw score is calculated by combining the weights and mapped to the risk level. Dynamic updates are performed throughout the entire life cycle of pre-event, during-event, and post-event.

[0014] The system generates decision recommendations by matching the preset credit granting strategy. When the credibility score changes or a high-priority risk signal is triggered, the weights are dynamically adjusted and the evaluation results are updated synchronously.

[0015] Preferably, the calculation of the comprehensive credibility score employs the following dynamic weight coupling algorithm:

[0016] First, the objective weights of the indicators are calculated based on the entropy weight method. Then, the subjective weights of the indicators are calculated based on the analytic hierarchy process. The final indicator weights are obtained by coupling the data source type coefficient λ, the business relevance coefficient μ, and the quality fluctuation sensitivity coefficient σ. The score is calculated according to Score∑4i=1ω1*xi, where xi is the quantitative value of accuracy, completeness, timeliness, and stability.

[0017] Preferably, the extraction of key risk signals specifically includes:

[0018] A multi-dimensional risk keyword library is constructed based on financial credit risk scenarios. Risk content is identified by precise matching and semantic similarity matching. Sentiment intensity is calculated through a pre-trained model in the financial field. Signals that are relevant to the publishing entity, authoritative from the source, and within the effective period are selected to generate a structured feature set. Signals with sentiment intensity ≥ 0.9 and source authority ≥ 0.8 are judged as high-priority risk signals.

[0019] Preferably, the custom risk control indicators specifically include:

[0020] It provides a visual formula editor that supports arithmetic, logical, and statistical functions. User-defined indicator calculation rule variables are mapped and associated with preprocessed data fields, and the system automatically performs indicator calculations and updates.

[0021] Preferably, the full-cycle dynamic update specifically includes:

[0022] The pre-credit scoring is calculated based on historical data and current static data; the in-credit scoring is updated according to a preset period or triggered events, which include at least one of the following: financial data updates, public opinion releases, and repayment anomalies; the post-credit scoring is supplemented by data on actual repayment performance within the credit period.

[0023] A financial risk control assessment system based on multi-source data fusion, used to implement a financial risk control assessment method based on multi-source data fusion, includes:

[0024] The data governance module is used to complete multi-source data access, cleaning, reliable quantification, hierarchical management, and risk signal extraction;

[0025] The indicator modeling module is used to support users in defining custom risk control indicators, building hierarchical risk control models, and managing versions.

[0026] The full-cycle scoring module is used to generate risk control feature matrices, calculate risk scores and map levels, and supports dynamic updates;

[0027] The intelligent decision-making module is used to match risk scores with credit granting strategies and generate decision recommendation reports;

[0028] The traceability and auditing module is used to store the entire operation log and evaluation results, and supports querying and tracing.

[0029] Preferably, the data governance module includes an API interface connection unit, a web crawler unit, a file import unit, a trusted quantification unit, and a risk signal extraction unit, wherein the trusted quantification unit has a built-in subjective-objective weight coupling algorithm, and the risk signal extraction unit integrates a pre-trained sentiment analysis model in the financial field.

[0030] Preferably, the indicator modeling module includes a visual formula editor, which allows users to edit indicator calculation rules by dragging and dropping or by using scripts, and the rule variables are automatically mapped to the fields of the data governance module.

[0031] Preferably, the full-cycle scoring module has a built-in dynamic weight adjustment engine. When the data source credibility score changes by ≥5 points or a high-priority risk signal is detected, the fusion weight and risk signal supplementary weight are automatically adjusted, and the score update is completed within 3 minutes.

[0032] A computer-readable storage medium having a computer program stored thereon, the program being executed by a processor to implement a financial risk control assessment method based on multi-source data fusion.

[0033] The technical effects and advantages of this invention are as follows:

[0034] 1. Accurate Measurement of Data Credibility: For the first time, accuracy, completeness, timeliness, and stability are transformed into calculable indicators. Combined with a dynamic weighting algorithm with multi-dimensional constraints, it solves the industry pain point of ambiguity in measuring the credibility of multi-source data. The response speed of data source weight adjustment is improved from the traditional monthly level to the hourly level.

[0035] 2. Strong scenario adaptability: Through the customizable indicators and model configuration functions, a single system can simultaneously cover multiple business scenarios such as personal credit, corporate credit, and supply chain finance, without the need for repeated development, and the model iteration cycle is shortened from several weeks to 1-3 days.

[0036] 3. Comprehensive Risk Capture: Low-reliability data is neither discarded directly nor arbitrarily used to interfere with the scoring. Through independent risk signal extraction and supplementary weight control, significant risk information such as debt defaults and regulatory penalties is retained while avoiding public opinion noise from lowering the accuracy of the scoring.

[0037] 4. End-to-end traceability and compliance: All operations from data collection, credibility calculation, model configuration to scoring decision are recorded and stored, supporting multi-dimensional traceability by customer, time, and scenario, fully complying with financial regulatory audit requirements. Attached Figure Description

[0038] Figure 1 This is a flowchart of an embodiment of the present invention; Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] Taking a city commercial bank that simultaneously conducts two types of business—personal business loans and credit lines for small and micro enterprises—as an example, the specific operational process of this invention is explained as follows:

[0041] Step 1: Data Access and Trust Measurement

[0042] The system synchronizes internal transaction records and credit records through API interfaces, pulls third-party credit and tax data daily, and collects corporate public opinion and industry dynamics through web crawlers; it unifies field naming to standard formats such as overdue_times (number of overdue payments) and debt_ratio (debt ratio), and removes abnormal transaction records other than 3σ.

[0043] The credibility of a certain third-party credit data source was calculated as follows: data matching rate 92%, field completeness rate 88%, data freshness 95%, and stability 90%. Combined with the fact that it is third-party data (type coefficient λ=0.6), business relevance of small and micro enterprises (μ=0.8), and fluctuation sensitivity σ=1.5, the final comprehensive credibility score was calculated to be 84 points, which was classified into the high credibility layer and assigned a fusion weight of 0.7.

[0044] Step 2: Customize model configuration

[0045] Risk control personnel build two models using a visual interface:

[0046] Personal business loan model: It consists of three sections: "Repayment ability (weight 40%), credit record (35%), and business stability (25%)". Among them, "Repayment ability" is related to indicators such as "average cash flow in the past 6 months" and "debt-to-income ratio". The calculation rule is defined by the formula editor as: Repayment ability score = (average cash flow / industry average) × 40 + (1 - debt-to-income ratio) × 60.

[0047] Credit model for small and micro enterprises: It consists of four sections: "financial status (35%), operational risk (30%), credit record (25%), and industry prospects (10%)". A separate sub-version for science and technology innovation enterprises is configured, and the weight of the "R&D investment ratio" indicator is increased from 5% to 15%.

[0048] The two models are each saved as version V1.0 and associated with the corresponding business scenarios.

[0049] Step 3: Full-cycle scoring execution

[0050] Pre-approval: When a company applies for credit, the system combines its internal financial reports, third-party credit reports, and public opinion data to calculate an original score of 78 points, which is mapped to Grade A, and matches a credit recommendation of "credit limit of 3 million and interest rate of 6%".

[0051] In-process update: The system detected new public opinion regarding the company’s “tax arrears announcement”, with an emotional intensity of 0.92 and a source authority of 0.85. This was determined to be a high-priority risk signal. The supplementary weight was increased from 0.1 to 0.2, and the recalculated score was 71 points, which was downgraded to BBB level. The system automatically triggered an alarm and suggested reducing the credit limit by 20%.

[0052] Post-mortem analysis: After the credit line expired, the system combined the company's repayment records and business changes to retrospectively score the accuracy of the rating, and simultaneously optimized the weighting rules of the "tax arrears announcement" signal into the keyword database.

[0053] Step 4: Retrospective Audit

[0054] All operation logs (including data source credibility calculation process, model configuration modification records, scoring details, and decision results) are linked and stored in the traceability audit module. Regulatory authorities can retrieve the entire lifecycle assessment chain of the enterprise to verify the compliance of risk control decisions.

[0055] The applicant further declares that while the above embodiments illustrate the implementation method and apparatus structure of the present invention, the present invention is not limited to the above-described embodiments, meaning that the present invention must rely on the above methods and structures to be implemented. Those skilled in the art should understand that any improvements to the present invention, equivalent substitutions for the selected implementation methods, additions to steps, and selections of specific methods all fall within the protection and disclosure scope of the present invention.

[0056] This invention is not limited to the above-described embodiments. All methods that employ similar structures and approaches to achieve the objectives of this invention are within the scope of protection of this invention.

Claims

1. A financial risk control assessment method based on multi-source data fusion, characterized in that, Including the following steps: Access to three types of data sources: internal data from financial institutions, third-party compliant data, and low-reliability supplementary data, and complete format unification and cleaning processes; The accuracy, completeness, timeliness, and stability of each data source are quantified by data matching rate, field completeness rate, data freshness, and data quality fluctuation coefficient, and a comprehensive credibility score is calculated by combining multi-dimensional constraint parameters. The data is divided into three layers—high, medium, and low—based on its overall credibility score, and fusion weights are assigned accordingly. Key risk signals are extracted from the low-credibility data. Respond to user operations by customizing risk control indicators and calculation rules, and construct a hierarchical risk control model that includes multiple assessment modules; The data features and risk signals of each layer are spliced ​​together, and the raw score is calculated by combining the weights and mapped to the risk level. Dynamic updates are performed throughout the entire life cycle of pre-event, during-event, and post-event. The system generates decision recommendations by matching the preset credit granting strategy. When the credibility score changes or a high-priority risk signal is triggered, the weights are dynamically adjusted and the evaluation results are updated synchronously.

2. The financial risk control assessment method based on multi-source data fusion according to claim 1, characterized in that, The calculation of the overall credibility score employs the following dynamic weight coupling algorithm: First, the objective weights of the indicators are calculated based on the entropy weight method. Then, the subjective weights of the indicators are calculated based on the analytic hierarchy process (AHP). Finally, the final indicator weights are obtained by coupling the data source type coefficient λ, the business relevance coefficient μ, and the quality fluctuation sensitivity coefficient σ. The weights are then calculated according to the Score∑. 4 i=1 ω 1* x i Calculate the score, where x i It is a quantitative value for accuracy, completeness, timeliness, and stability.

3. The financial risk control assessment method based on multi-source data fusion according to claim 1, characterized in that, The extraction of key risk signals specifically includes: A multi-dimensional risk keyword library is constructed based on financial credit risk scenarios. Risk content is identified by precise matching and semantic similarity matching. Sentiment intensity is calculated through a pre-trained model in the financial field. Signals that are relevant to the publishing entity, authoritative from the source, and within the effective period are selected to generate a structured feature set. Signals with sentiment intensity ≥ 0.9 and source authority ≥ 0.8 are judged as high-priority risk signals.

4. The financial risk control assessment method based on multi-source data fusion according to claim 1, characterized in that, The custom risk control indicators specifically include: It provides a visual formula editor that supports arithmetic, logical, and statistical functions. User-defined indicator calculation rule variables are mapped and associated with preprocessed data fields, and the system automatically performs indicator calculations and updates.

5. The financial risk control assessment method based on multi-source data fusion according to claim 1, characterized in that, The full-cycle dynamic update specifically includes: The pre-credit scoring is calculated based on historical data and current static data; the in-credit scoring is updated according to a preset period or triggered events, which include at least one of the following: financial data updates, public opinion releases, and repayment anomalies; the post-credit scoring is supplemented by data on actual repayment performance within the credit period.

6. A financial risk control assessment system based on multi-source data fusion, used to implement the financial risk control assessment method based on multi-source data fusion as described in any one of claims 1-5, characterized in that, include: The data governance module is used to complete multi-source data access, cleaning, reliable quantification, hierarchical management, and risk signal extraction; The indicator modeling module is used to support users in defining custom risk control indicators, building hierarchical risk control models, and managing versions. The full-cycle scoring module is used to generate risk control feature matrices, calculate risk scores and map levels, and supports dynamic updates; The intelligent decision-making module is used to match risk scores with credit granting strategies and generate decision recommendation reports; The traceability and auditing module is used to store the entire operation log and evaluation results, and supports querying and tracing.

7. A financial risk control assessment system based on multi-source data fusion according to claim 6, characterized in that, The data governance module includes an API interface connection unit, a web crawler unit, a file import unit, a trusted quantification unit, and a risk signal extraction unit. The trusted quantification unit has a built-in subjective and objective weight coupling algorithm, and the risk signal extraction unit integrates a pre-trained sentiment analysis model in the financial field.

8. A financial risk control assessment system based on multi-source data fusion according to claim 6, characterized in that, The indicator modeling module includes a visual formula editor, which allows users to edit indicator calculation rules by dragging and dropping or using scripts. The rule variables are automatically mapped to the fields of the data governance module.

9. A financial risk control assessment system based on multi-source data fusion according to claim 6, characterized in that, The full-cycle scoring module has a built-in dynamic weight adjustment engine. When the data source credibility score changes by ≥5 points or a high-priority risk signal is detected, it automatically adjusts the fusion weight and the risk signal supplementary weight, and completes the score update within 3 minutes.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the financial risk control assessment method based on multi-source data fusion as described in any one of claims 1-5.

Images