A method for mining credit report features
Through automated feature mining methods, the People's Bank of China's credit report is processed to generate comprehensive and effective features, and the problem of mining the characteristics of the People's Bank of China's credit report in the existing technology requires a lot of manpower and material resources, and the efficient mining and interpretation of information is achieved. feature generation.
Patent Information
- Application Number
- CN202111410257.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-11-23
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2041-11-23
AI Technical Summary
In the prior art, the mining of the characteristics of the People's Bank of China's credit report requires a lot of manpower and material resources, and the information cannot be fully explored.
Adopt automated feature mining methods, including obtaining the People's Bank of China's credit report and automatically parsing it into tabular data, and then generating derivative features through automated feature mining procedures, and filtering the characteristics that meet the conditions based on preset thresholds, providing feature interpretation, and ensuring comprehensive information Dig.
It saves a lot of manpower and material resources, and the generated characteristics are comprehensive and effective, with strong interpretability, and is suitable for a variety of data structures.
Smart Images

Figure CN114140219B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of credit risk technology, and in particular relates to a method for mining credit report features. Background Art
[0002] The People's Bank of China's credit report is divided into personal credit report and corporate credit report. It is a record of personal credit information issued by the Credit Reference Center of the People's Bank of China and is used to inquire about the social credit of individuals or enterprises.
[0003] Credit reports are divided into three categories: basic personal information, credit transaction information and other information.
[0004] The PBOC credit report records the historical information of credit transactions between customers and banks and other financial institutions, and is a key data source for most financial institutions to assess the risk level of loan applicants. The PBOC credit report contains rich information, and full feature mining can effectively improve the ability of financial institutions to distinguish between good and bad customers.
[0005] At present, many financial institutions also do some feature mining when using the PBOC credit report. However, due to the complex structure and high information density of the PBOC credit report, it often requires a lot of manpower and material resources, and a lot of information cannot be fully mined. Summary of the invention
[0006] The present invention discloses a method for mining the features of a credit report, and is intended to solve the technical problem mentioned in the background technology that currently mining the features of the credit report of the People's Bank of China requires a lot of manpower and material resources, and a lot of information cannot be fully mined.
[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is as follows:
[0008] A method for mining credit report features comprises the following steps:
[0009] Step 1: Obtain the credit report from the People's Bank of China, automatically parse the credit report, and process it into table data;
[0010] Step 2: Group the tabular data obtained in step 1 according to individuals, and respectively establish corresponding array structures for storing the data after the individual groups are formed, and then read the data in the array structures through an automated feature mining program, perform feature mining on the tabular data after the individual groups are formed, and obtain derived features and feature interpretations corresponding to the derived features;
[0011] Step 3: Based on a preset threshold, the derived features generated in step 2 are screened to select the derived features that meet the conditions.
[0012] The feature mining of the present invention is carried out automatically, which saves a lot of manpower and material resources; and can generate feature explanations of corresponding derived features, indicating the calculation path of each feature, and has strong interpretability.
[0013] Preferably, step 1 further comprises performing sample screening on the obtained table data to screen out table data within the past 30 days or the past 60 days.
[0014] Preferably, after the table data is grouped according to individuals in step 2, it is also necessary to group them according to categories, and establish corresponding array structures for storing the data after category grouping.
[0015] Furthermore, before using the automated feature mining program to perform feature mining on the table data after the individual groups in step 2, feature conversion is performed on important information in the table that can reflect the customer's default risk.
[0016] The present invention performs feature conversion on important information in the table that can reflect the customer's default risk, then performs feature extraction, and finally generates derived features to avoid missing certain important default risk information; for example: converting the timestamp into a time difference can be used to characterize the time interval pattern between several queries by the customer within a period of time, which is also important information reflecting the customer's default risk.
[0017] Preferably, the table data in step 1 includes basic information, public information, account summary information, query information, credit limit usage information and repayment information.
[0018] Preferably, in step 2, different calculation functions are used to perform statistical aggregation on continuous variables and discrete variables to generate derived features in batches; and corresponding feature explanations are generated according to the calculation paths and calculation functions of the derived features generated in batches.
[0019] The corresponding calculation functions are used to summarize and count the different calculation paths under different data dimensions, and derivative features are generated in batches to ensure that all dimensions of the People's Bank of China's credit report can be fully mined and the mined features are comprehensive and effective.
[0020] Preferably, the calculation path includes: limiting data time, classifying and aggregating by various categories, and variable conversion.
[0021] Preferably, the continuous variables include: sum, minimum, maximum, mean, standard deviation, skewness and other custom functions; the discrete variables include count, proportion and other custom functions.
[0022] Through the customizable functions, the application scenarios of the present invention are not limited to the credit report of the People's Bank of China, but can also be applied to other similar data structures for feature mining.
[0023] Preferably, the threshold in step 3 is to retain the derived features with a feature missing rate less than 90% and a variance greater than 0.
[0024] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:
[0025] 1. The feature mining of the present invention is carried out automatically, which saves a lot of manpower and material resources; and can generate feature explanations of corresponding derived features, indicating the calculation path of each feature, and has strong interpretability.
[0026] 2. Use corresponding calculation functions to summarize and count different calculation paths under different data dimensions, and generate derivative features in batches to ensure that all dimensions of the People's Bank of China's credit report can be fully mined and the mined features are comprehensive and effective.
[0027] 3. Through the customizable functions, the application scenarios of the present invention are not limited to the credit report of the People's Bank of China, but can also be applied to other similar data structures for feature mining. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] The present invention will now be described by way of example with reference to the accompanying drawings, in which:
[0029] Figure 1 It is a processing logic block diagram of the present invention. DETAILED DESCRIPTION
[0030] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application usually described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.
[0031] The following is combined with Figure 1 Describe the embodiments of the present invention in detail;
[0032] A method for mining credit report features comprises the following steps:
[0033] Step 1: Obtain the credit report of the People's Bank of China, automatically parse the credit report, process it into table data, and perform sample screening on the obtained table data to select table data within the past 30 days or the past 60 days; the table data includes basic information, public information, account summary information, query information, credit limit usage information and repayment information;
[0034] Step 2: Group the table data obtained in step 1 by individuals, and establish corresponding array structures to store the data after individual grouping, group by category (such as query records of banking institutions and non-banking institutions), and establish corresponding array structures to store the data after category grouping, and perform feature conversion on the important information in the data that can reflect the customer's default risk (for example: converting the timestamp to the time difference can be used to characterize the time interval between several queries of the customer within a period of time, which is also important information reflecting the customer's default risk), and then read the data in the array structure through the automated feature mining program, use different calculation functions to perform statistical aggregation on continuous variables and discrete variables, and generate derived features in batches; and generate corresponding feature explanations based on the calculation path and calculation function of the derived features generated in batches. Use the corresponding calculation function to summarize and count different calculation paths under different data dimensions, and generate derived features in batches to ensure that all dimensions of the credit report of the People's Bank of China can be fully mined, and the mined features are comprehensive and effective.
[0035] The calculation path includes: limiting data time, classifying and aggregating by various categories, and variable conversion.
[0036] The continuous variables include: sum, minimum, maximum, mean, standard deviation, skewness and other custom functions; the discrete variables include count, proportion and other custom functions.
[0037] Through the customizable functions, the application scenarios of the present invention are not limited to the credit report of the People's Bank of China, but can also be applied to other similar data structures for feature mining.
[0038] Step 3: Based on a preset threshold (retaining derived features with a feature missing rate less than 90% and a variance greater than 0), screen the derived features generated in step 2 to select derived features that meet the conditions.
[0039] The feature mining of the present invention is carried out automatically, which saves a lot of manpower and material resources; and can generate feature explanations of corresponding derived features, indicating the calculation path of each feature, and has strong interpretability.
[0040] The present invention performs feature conversion on important information in the table that can reflect the customer's default risk, then performs feature extraction, and finally generates derived features to avoid missing certain important default risk information; for example: converting the timestamp into a time difference can be used to characterize the time interval pattern between several queries by the customer within a period of time, which is also important information reflecting the customer's default risk.
[0041] The present invention uses various combinations of data dimensions, calculation paths, and calculation functions, and the number of derived features automatically generated through the above steps is more than 10,000, which fully mines the information of each part of the credit report. The whole process is automatic, and the generated features have corresponding explanations, saving a lot of time and manpower.
[0042] The feature mining of the present invention is carried out automatically, which saves a lot of manpower and material resources; and can generate feature explanations of corresponding derived features, indicating the calculation path of each feature, and has strong interpretability.
[0043] The above-mentioned embodiments only express the specific implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the protection scope of the present application. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the technical solution concept of the present application, and these all belong to the protection scope of the present application.
Claims
1. A method for mining credit report features, characterized in that: The following steps are involved: Step 1: Obtain the credit report from the People's Bank of China, automatically parse the credit report, and process it into table data; Step 2: Group the tabular data obtained in step 1 according to individuals, and respectively establish corresponding array structures for storing the data after the individual groups are formed, and then read the data in the array structures through an automated feature mining program, perform feature mining on the tabular data after the individual groups are formed, and obtain derived features and feature interpretations corresponding to the derived features; Step 3: Based on a preset threshold, the derived features generated in step 2 are screened to select the derived features that meet the conditions; After the table data is grouped by individuals in step 2, it is also necessary to group them by category, and establish corresponding array structures for storing the data after category grouping; Before using the automated feature mining program to perform feature mining on the table data after the individual groups in step 2, feature conversion is performed on the important information in the table that can reflect the customer's default risk; In step 2, different calculation functions are used to perform statistical aggregation on continuous variables and discrete variables, and derivative features are generated in batches; And generate corresponding feature explanations based on the calculation paths and calculation functions of the batch-generated derivative features.
2. A method for mining credit report features according to claim 1, characterized in that: The step 1 also includes performing sample screening on the obtained table data to screen out table data within the past 30 days or the past 60 days.
3. The method for mining credit report features according to claim 1, characterized in that: The table data in step 1 includes basic information, public information, account summary information, query information, credit limit usage information and repayment information.
4. The method for mining credit report features according to claim 1, characterized in that: The calculation path includes: limiting data time, classifying and aggregating by various categories, and variable conversion.
5. The method for mining credit report features according to claim 1, characterized in that: The continuous variables include: sum, minimum, maximum, mean, standard deviation, skewness and other custom functions; the discrete variables include count, proportion and other custom functions.
6. The method for mining credit report features according to claim 1, characterized in that: The threshold in step 3 is to retain the derived features whose feature missing rate is less than 90% and whose variance is greater than 0.
