Financial credit risk identification method and system

Abstract

The invention discloses a financial credit risk identification method and system. The method comprises the following steps: acquiring information of borrowers from a plurality of data sources; cleaning, duplicate removal, missing value filling and standardization processing are carried out on the collected data; constructing a multi-dimensional feature set based on the preprocessed data; training the feature set by using a machine learning algorithm in combination with historical credit default cases to generate a credit risk prediction model; inputting data of a to-be-evaluated borrower into the trained credit risk prediction model, and outputting a credit risk score and a potential risk point prompt; credit behavior changes of borrowers are tracked continuously, and the risk model is updated regularly to adapt to market and economic environment changes. According to the method, comprehensive data is obtained through multi-data source fusion, the limitation of traditional single data source assessment is broken, and risk assessment is more comprehensive and accurate; through refined data preprocessing and scientific feature construction, deep information of data can be mined, and high-quality input is provided for model training.

Description

Technical Field

[0001] The present invention relates to the field of financial technology, and specifically refers to a method and system for identifying financial credit risks. Background Art

[0002] In the financial field, credit business is one of the core businesses of financial institutions such as banks. However, with the increasing complexity and diversification of the financial market, credit risks have also increased. Accurately identifying and assessing credit risks is of crucial significance for financial institutions to ensure the safety of funds and optimize resource allocation. Traditional credit risk assessment methods often rely on a single credit score or limited financial indicators, making it difficult to comprehensively and accurately reflect the true risk status of borrowers, and unable to adapt to the dynamic changes of the market and economic environment in a timely manner.

[0003] Therefore, there is an urgent need for a more efficient, accurate and adaptable method and system for identifying financial credit risks. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for identifying financial credit risks to solve the deficiencies in the background art.

[0005] To achieve the above purpose, the present invention provides the following technical solutions: A method for identifying financial credit risks includes the following steps.

[0006] S1. Obtain the historical transaction records, credit scores, financial status, personal basic information and industry-related data of the borrower from multiple data sources.

[0007] S2. Clean, de-duplicate, fill in missing values and standardize the collected data.

[0008] S3. Based on the preprocessed data, construct a multi-dimensional feature set including repayment behavior characteristics, debt ratio characteristics, income stability characteristics, and social credit environment characteristics.

[0009] S4. Use machine learning algorithms, combined with historical credit default cases, to train the feature set to generate a credit risk prediction model.

[0010] S5. Input the data of the borrower to be evaluated into the trained credit risk prediction model, and output the credit risk score and potential risk point tips.

[0011] S6. Continuously track the changes in the borrower's credit behavior, and regularly update the risk model to adapt to the changes in the market and economic environment.

[0012] Further, in step S1, obtain the information of the data source through API interfaces, web crawler technology or direct data exchange methods.

[0013] Further, in step S1, the data sources include

[0014] The internal banking transaction system, which obtains the detailed historical transaction records of the borrower in this bank, including the repayment time, amount, and overdue situation of each loan;

[0015] The database of professional credit rating agencies, which accurately extracts the latest authoritative credit scores of borrowers through an authorized API interface, and the scoring basis covers multi-dimensional indicators such as past borrowing performance and credit inquiry frequency;

[0016] The filing information of the tax department and the industrial and commercial department, which obtains the balance sheet and income statement data of the enterprise, as well as the individual tax declaration records, and at the same time collects the borrower's age, occupation, and educational background information;

[0017] The industry dynamic data released by industry associations and research institutions, which obtains the industry-related data such as the overall growth trend, competition pattern, and policy and regulation changes in the industry where the borrower is located.

[0018] Further, in step S2, regular expressions and a predefined rule library are used to identify and remove error characters and garbled codes in the transaction records, and filter out exactly the same duplicate data records;

[0019] For missing values, numerical data is filled based on the mean, median of the same type of borrowers in this field or the predicted value based on machine learning;

[0020] The standardization process uses the Z-score standardization method to uniformly map data of different magnitudes into a standard normal distribution space with a mean of 0 and a standard deviation of 1.

[0021] Further, in step S3, the construction of the repayment behavior characteristics introduces volatility indicators of the repayment time series, and measures the borrower's repayment stability by calculating the standard deviation and coefficient of variation;

[0022] The debt ratio characteristics include the ratio of total debt to total assets and the ratio of current liabilities to current assets;

[0023] The income stability characteristics combine the diversity of income sources and the income growth trend, and use a sliding window to calculate the month-on-month growth rate and coefficient of variation of the income in the past year;

[0024] The social credit environment characteristics consider the credit culture atmosphere in the region where the borrower is located, and statistically include the average credit score of the region and the density index of dishonest executors.

[0025] Further, in step S3, methods such as principal component analysis and cluster analysis are used for feature dimensionality reduction to improve the model operation efficiency.

[0026] Furthermore, in step S4, the model parameters are optimized by using cross-validation and grid search methods to ensure the generalization ability of the model.

[0027] The present invention also provides a financial credit risk identification system, comprising

[0028] Data collection module: collects borrowers’ historical transaction records, credit scores, financial status, basic personal information and industry-related data from multiple data sources through API interfaces, crawler technology or direct data exchange;

[0029] Data preprocessing module: used to identify and remove incorrect characters and garbled characters in transaction records, and filter out duplicate data records that are identical;

[0030] Feature construction module: used to construct a multi-dimensional feature set including repayment behavior characteristics, debt ratio characteristics, income stability characteristics, and social credit environment characteristics;

[0031] Model training module: It is used to train the feature set based on historical credit default cases, and optimize the model parameters using cross-validation and grid search methods to ensure the generalization ability of the model, thereby generating a credit risk prediction model.

[0032] Risk assessment module: used to input the data of the borrower to be assessed into the trained credit risk prediction model, and output the credit risk score and potential risk point prompts.

[0033] Model update module: Continuously track changes in borrowers' credit behavior, regularly collect relevant new data, and re-drive the data collection, preprocessing, and feature construction processes based on new data and changes in the market and economic environment to update the risk model to adapt it to changes in the market and economic environment.

[0034] The advantages of the present invention compared with the prior art are: on the one hand, the present invention obtains comprehensive data through the fusion of multiple data sources, breaking the limitations of traditional single data source assessment and making risk assessment more comprehensive and accurate; on the other hand, refined data preprocessing and scientific feature construction can mine deep-level information of the data and provide high-quality input for model training; furthermore, by using advanced machine learning algorithms and parameter optimization strategies, the generated credit risk prediction model has strong generalization ability and can adapt to different types of borrowers; finally, the continuous model update mechanism ensures that the system can dynamically respond to changes in the market and economic environment, effectively reduce financial credit risks, and improve the risk management level of financial institutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a flow chart of a financial credit risk identification method of the present invention.

[0036] Figure 2This is the system block diagram of a financial credit risk identification system of the present invention. Specific implementation manners

[0037] The following further elaborates on a financial credit risk identification method and system of the present invention with reference to the accompanying drawings.

[0038] Combined with the attached Figure 1-2 , the present invention is introduced in detail.

[0039] A financial credit risk identification method includes the following steps:

[0040] Step S1: Data acquisition

[0041] Obtain the historical transaction records, credit scores, financial conditions, personal basic information, and industry-related data of borrowers from multiple data sources. Specifically, obtain the information of the data sources through API interfaces, web crawler technologies, or direct data exchange methods. Among them, the data sources cover the internal transaction systems of banks, and can obtain the detailed historical transaction records of borrowers in the bank, such as the repayment time, amount, and overdue situation of each loan; the databases of professional credit rating agencies, and accurately extract the latest authoritative credit scores of borrowers through authorized API interfaces. The scoring basis covers multi-dimensional indicators such as past borrowing performance and credit query frequency; the filing information of the tax department and the industrial and commercial department can obtain the balance sheet and income statement data of enterprises, as well as personal tax return records, and at the same time collect information such as the age, occupation, and education level of borrowers; the industry dynamic data released by industry associations and research institutions is used to obtain industry-related data such as the overall growth trend, competition pattern, and policy and regulation changes in the industry where the borrowers are located.

[0042] Step S2: Data preprocessing

[0043] Clean, de-duplicate, fill in missing values, and standardize the collected data. Use regular expressions and predefined rule libraries to identify and remove error characters and garbled codes in transaction records, and filter out exactly the same data records that are repeatedly entered; for missing values, numerical data is filled based on the mean, median of the same type of borrowers in this field, or prediction values based on machine learning; the standardization process uses the Z-score standardization method to uniformly map data of different magnitudes to a standard normal distribution space with a mean of 0 and a standard deviation of 1.

[0044] Step S3: Feature construction

[0045] Based on the preprocessed data, construct a multi-dimensional feature set including repayment behavior characteristics, debt ratio characteristics, income stability characteristics, and social credit environment characteristics. The construction of the repayment behavior characteristics introduces volatility indicators of the repayment time series, and measures the borrower's repayment stability by calculating the standard deviation and coefficient of variation; the debt ratio characteristics include the ratio of total debt to total assets and the ratio of current liabilities to current assets; the income stability characteristics combine the diversity of income sources and the income growth trend, and use a sliding window to calculate the month-on-month growth rate and coefficient of variation of income in the past year; the social credit environment characteristics consider the credit culture atmosphere in the borrower's region, and statistically incorporate the average credit score and the density of dishonest executors in the region. In addition, feature dimensionality reduction can also be performed through principal component analysis and clustering analysis methods to improve the model operation efficiency.

[0046] Step S4: Model training

[0047] Use machine learning algorithms and combine historical credit default cases to train the feature set to generate a credit risk prediction model. In this process, use cross-validation and grid search methods to optimize the model parameters to ensure the generalization ability of the model.

[0048] Step S5: Risk assessment

[0049] Input the data of the borrower to be evaluated into the trained credit risk prediction model, and output the credit risk score and potential risk point tips.

[0050] Step S6: Model update

[0051] Continuously track the changes in the borrower's credit behavior, and regularly update the risk model to adapt to market and economic environment changes. That is, continuously track the changes in the borrower's credit behavior, regularly collect relevant new data, and re-drive the data collection, preprocessing, and feature construction processes based on the new data and market and economic environment changes to update the risk model to make it adapt to market and economic environment changes.

[0052] To implement the above method, the present invention provides a financial credit risk identification system, including

[0053] Data collection module: Through API interfaces, web crawler technology, or direct data exchange, collect the borrower's historical transaction records, credit scores, financial status, personal basic information, and industry-related data from multiple data sources, corresponding to the data acquisition step in the above method, to ensure comprehensive data collection.

[0054] Data preprocessing module: Used to identify and remove error characters and garbled characters in the transaction records, filter out exactly the same data records that are repeatedly entered, and complete the preliminary purification of the data to provide a high-quality data basis for subsequent analysis, corresponding to the data preprocessing step in the method.

[0055] Feature construction module: It is used to construct a multi-dimensional feature set including repayment behavior features, debt ratio features, income stability features, and social credit environment features, convert the preprocessed data into targeted feature information to assist in accurate risk assessment, and cooperate with the feature construction step in the method.

[0056] Model training module: It is used to train the feature set in combination with historical credit default cases, and optimize the model parameters by using cross-validation and grid search methods to ensure the generalization ability of the model, and then generate a credit risk prediction model. It is the key module to realize the construction and optimization of the risk prediction model, and matches the model training step in the method.

[0057] Risk assessment module: It is used to input the data of the borrower to be evaluated into the trained credit risk prediction model, output the credit risk score and prompt of potential risk points, and realize the quantitative assessment of the borrower's real-time risk, corresponding to the risk assessment step in the method.

[0058] Model update module: Continuously track the changes in the credit behavior of borrowers, regularly collect relevant new data, and re-drive the data collection, preprocessing, and feature construction processes according to the new data and changes in the market and economic environment to update the risk model, so that it can adapt to the changes in the market and economic environment, and ensure the timeliness and adaptability of the risk identification system, which fits the model update step in the method.

[0059] The specific implementation process of a financial credit risk identification method and system of the present invention is as follows:

[0060] In practical applications, financial institutions first start the data collection module, widely collect borrower information from various data sources according to the set API interface call rules, crawler strategies, and data exchange protocols. After the collection is completed, the data automatically flows into the data preprocessing module, and the data is cleaned by using the built-in regular expressions and rule libraries to remove error characters and garbled codes. At the same time, the duplicate data is filtered out by using the deduplication algorithm. For missing values, the system selects the mean or median of borrowers of the same type according to the preset filling strategy, or starts a machine learning prediction model to generate filling values, and performs Z-score standardization processing on all data.

[0061] The preprocessed data enters the feature construction module, which calculates the volatility index of repayment behavior, debt ratio, income stability index, and social credit environment quantification index according to the established feature construction algorithm. At the same time, it can call the principal component analysis or clustering analysis tool as needed for feature dimension reduction to optimize the feature set.

[0062] Next, the model training module receives the feature set, combines it with the historical credit default case library, and selects a suitable machine learning algorithm (such as decision tree, neural network, etc.) to train the model. During the training process, the model parameters are continuously adjusted through cross-validation and grid search until a credit risk prediction model with good generalization ability is generated.

[0063] When it is necessary to evaluate the risk of a borrower, the risk assessment module inputs the data of the borrower to be evaluated into the trained model, and quickly outputs the credit risk score and detailed potential risk point prompts for the reference of credit personnel in making decisions.

[0064] In daily operations, the model update module continuously monitors the credit behavior of borrowers, regularly (such as monthly or quarterly) re-collects data, and drives the entire process to update cyclically to ensure that the risk identification system always keeps up with the changes in the market and economic environment, and escorts the financial credit business.

[0065] The above describes the present invention and its implementation manners. Such a description is not restrictive. What is shown in the drawings is only one of the implementation manners of the present invention, and the actual structure is not limited thereto. All in all, if those of ordinary skill in the art are inspired by it and design similar structural manners and embodiments without creative efforts without departing from the purpose of the present invention, they shall fall within the protection scope of the present invention.

Claims

1. A method for identifying financial credit risk, characterized in that: The following steps are included: S1. Obtain the borrower's historical transaction records, credit scores, financial status, basic personal information and industry-related data from multiple data sources; S2. Clean, remove duplicates, fill missing values ​​and standardize the collected data; S3. Based on the preprocessed data, a multi-dimensional feature set including repayment behavior characteristics, debt ratio characteristics, income stability characteristics, and social credit environment characteristics is constructed; S4. Use machine learning algorithms and historical credit default cases to train feature sets and generate a credit risk prediction model; S5. Input the data of the borrower to be evaluated into the trained credit risk prediction model, and output the credit risk score and potential risk point prompts; S6. Continue to track changes in borrowers’ credit behavior and regularly update risk models to adapt to changes in the market and economic environment.

2. A financial credit risk identification method according to claim 1, characterized in that: In step S1, the information of the data source is obtained through an API interface, crawler technology or direct data exchange.

3. A financial credit risk identification method according to claim 2, characterized in that: In step S1, the data source includes The bank's internal transaction system obtains the borrower's detailed historical transaction records in the bank, including the repayment time, amount, and overdue status of each loan; The database of professional credit rating agencies can accurately extract the latest authoritative credit scores of borrowers through authorized API interfaces. The scoring is based on multi-dimensional indicators including past loan performance and credit inquiry frequency. The tax department and the industrial and commercial department record information, obtain the company's balance sheet, income statement data, and personal tax return records, and collect the borrower's age, occupation, and education information; Industry dynamics data released by industry associations and research institutions to obtain industry-related data on the overall growth trend, competitive landscape, and policy and regulatory changes in the borrower's industry.

4. A financial credit risk identification method according to claim 3, characterized in that: In step S2, regular expressions and predefined rule bases are used to identify and remove incorrect characters and garbled characters in transaction records, and to filter out duplicate data records that are identical; For missing values, numerical data is filled based on the mean, median, or machine learning-based predicted value of the field for borrowers of the same type; The standardization process uses the Z-score standardization method to uniformly map data of different magnitudes to a standard normal distribution space with a mean of 0 and a standard deviation of 1.

5. A financial credit risk identification method according to claim 4, characterized in that: In step S3, the repayment behavior feature construction introduces the volatility index of the repayment time series, and measures the repayment stability of the borrower by calculating the standard deviation and the coefficient of variation; The debt ratio characteristics include the ratio of total liabilities to total assets and the ratio of current liabilities to current assets; The income stability feature combines the diversity of income sources and income growth trends, and uses a sliding window to calculate the month-on-month growth rate and coefficient of variation of income in the past year; The social credit environment characteristics take into account the credit culture atmosphere in the borrower's region, and include quantitative indicators such as the average credit score of the region and the density of dishonest debtors.

6. A financial credit risk identification method according to claim 5, characterized in that: In step S3, feature dimensionality reduction is performed through principal component analysis and cluster analysis to improve the model operation efficiency.

7. A financial credit risk identification method according to claim 6, characterized in that: In step S4, the model parameters are optimized by using cross-validation and grid search methods to ensure the generalization ability of the model.

8. A financial credit risk identification system, characterized by: include Data collection module: collects borrowers’ historical transaction records, credit scores, financial status, basic personal information and industry-related data from multiple data sources through API interfaces, crawler technology or direct data exchange; Data preprocessing module: used to identify and remove incorrect characters and garbled characters in transaction records, and filter out duplicate data records that are identical; Feature construction module: used to construct a multi-dimensional feature set including repayment behavior characteristics, debt ratio characteristics, income stability characteristics, and social credit environment characteristics; Model training module: It is used to train the feature set based on historical credit default cases, and optimize the model parameters using cross-validation and grid search methods to ensure the generalization ability of the model, thereby generating a credit risk prediction model. Risk assessment module: used to input the data of the borrower to be assessed into the trained credit risk prediction model, and output the credit risk score and potential risk point prompts. Model update module: Continuously track changes in borrowers' credit behavior, regularly collect relevant new data, and re-drive the data collection, preprocessing, and feature construction processes based on new data and changes in the market and economic environment to update the risk model to adapt it to changes in the market and economic environment.

Images