Method and device for determining credit score of loan user and storage medium

Abstract

The invention discloses a credit score determination method and device of a loan user and a storage medium, and relates to the field of financial science and technology. The method comprises the steps that under the condition that user authorization is obtained, N types of financial data of a loan user are obtained, and N is an integer larger than 1; when it is detected that missing values and / or abnormal values exist in the i-th type of financial data in the N types of financial data, mathematical statistical data of the i-th type of financial data is calculated; correcting missing values and / or abnormal values in the i-th type of financial data according to the mathematical statistical data, and performing feature extraction on the corrected i-th type of financial data to obtain a group of feature sequences corresponding to the i-th type of financial data; and determining a target credit score of the loan user according to the N groups of feature sequences corresponding to the N types of financial data. According to the method and the device, the technical problem of low accuracy of determining the credit score of the user due to excessive types of user data in the prior art is solved.

Description

Technical Field

[0001] The present application relates to the field of financial technology, and specifically, to a method, device and storage medium for determining the credit score of a loan user. Background Art

[0002] With the continuous development of the economy and society, the products and services of the financial industry are facing many challenges while improving efficiency and serving the society. The most prominent one is how to effectively prevent and control risks. In the bank loan business, personal loan credit assessment is one of the important measures to prevent and control risks. Personal loan credit assessment can effectively identify customer default risks, prevent credit risks in advance, and reduce credit risks.

[0003] At present, banks mainly use manual review for personal loan credit assessment, that is, auditors manually review the loan application materials submitted by customers. Although manual review is low-cost, it cannot cope with the massive growth in personal loan demand and leads to low efficiency of bank loan business. In addition, due to factors such as inconsistent review standards and subjective judgment bias, the fairness and accuracy of the review results cannot be guaranteed. For this reason, banks have adopted a personal loan credit assessment system that is automated and assisted by manual review.

[0004] At present, banks mainly use rule-based personal loan credit assessment models, which review the loan application materials submitted by customers according to the rules in the rule base by building a rule base and a rule engine. Although this personal loan credit assessment model has low cost, it often fails to effectively identify customer default risks when processing large-scale and multi-category user data, resulting in technical problems such as low accuracy in determining user credit scores.

[0005] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0006] The embodiments of the present application provide a method, device and storage medium for determining the credit score of a loan user, so as to at least solve the technical problem in the prior art of low accuracy in determining the credit score of a user due to too many types of user data.

[0007] According to one aspect of an embodiment of the present application, a method for determining a credit score of a loan user is provided, comprising: obtaining N types of financial data of the loan user with the user's authorization, wherein N is an integer greater than 1; calculating mathematical statistical data of the i-th type of financial data when missing values ​​and / or abnormal values ​​are detected in the i-th type of financial data among the N types of financial data, wherein the i-th type of financial data is any type of financial data among the N types of financial data; correcting the missing values ​​and / or abnormal values ​​in the i-th type of financial data according to the mathematical statistical data, and extracting features from the corrected i-th type of financial data to obtain a set of feature sequences corresponding to the i-th type of financial data; and determining a target credit score of the loan user according to the N groups of feature sequences corresponding to the N types of financial data.

[0008] Optionally, when missing values ​​and / or outliers are detected in the i-th category of financial data among N categories of financial data, mathematical statistical data of the i-th category of financial data are calculated, including: when missing values ​​are detected in the i-th category of financial data, the average value of the i-th category of financial data is calculated; and / or; when outliers are detected in the i-th category of financial data, the average value and standard deviation of the i-th category of financial data are calculated.

[0009] Optionally, the missing values ​​and / or outliers in the i-th category of financial data are corrected according to mathematical statistical data, including: when missing values ​​are detected in the i-th category of financial data, using the average value of the i-th category of financial data to fill the missing values ​​in the i-th category of financial data; and / or; when outliers are detected in the i-th category of financial data, calculating the difference between the average value of the i-th category of financial data and the outlier, and performing any of the following correction operations on the outlier according to the absolute value of the difference and the standard deviation of the i-th category of financial data: retaining the outlier, deleting the outlier, and modifying the outlier.

[0010] Optionally, the outlier is modified or deleted according to the absolute value of the difference and the standard deviation of the i-th category financial data, including: calculating the product of the standard deviation and a preset coefficient to obtain a first value, wherein the preset coefficient is used to amplify the standard deviation or reduce the standard deviation; when it is detected that the absolute value of the difference is greater than the first value, deleting the outlier or modifying the outlier, wherein the absolute value of the difference between the modified outlier and the average value of the i-th category financial data is less than or equal to the first value; when it is detected that the absolute value of the difference is less than or equal to the first value, retaining the outlier.

[0011] Optionally, the credit score of the loan user is determined based on N groups of feature sequences corresponding to N types of financial data, including: inputting the N groups of feature sequences into N model branches respectively, wherein the j-th model branch among the N model branches is used to process the j-th group of feature sequences among the N groups of feature sequences, and the j-th type of financial data corresponding to the j-th group of feature sequences is the same type of data as the training data of the j-th model branch; determining an initial credit score of the loan user by each model branch based on a set of feature sequences input and in combination with prior knowledge of the model branch during the training process, wherein the N model branches generate N initial credit scores based on the N groups of feature sequences; and calculating the average of the N initial credit scores to obtain the target credit score of the loan user.

[0012] Optionally, the credit score of the loan user is determined based on N groups of feature sequences corresponding to N types of financial data, including: classifying the N groups of feature sequences into a first type of feature sequence and a second type of feature sequence, wherein the number of features in the first type of feature sequence is positively correlated with the credit score of the loan user, and the number of features in the second type of feature sequence is negatively correlated with the credit score of the loan user; inputting the first type of feature sequence into a first prediction model, and determining a first credit score based on the first type of feature sequence using prior knowledge learned by the first prediction model in a model training phase, wherein the first credit score is a value greater than 0, and the training data of the first prediction model is historical loan usage. The first prediction model is based on the credit behavior data of historical loan users, and the training label of the first prediction model is the credit score set for the credit behavior data of historical loan users; the second feature sequence is input into the second prediction model, and the second credit score is determined according to the second feature sequence through the prior knowledge learned by the second prediction model in the model training stage, wherein the second credit score is a value less than 0, the training data of the second prediction model is the dishonest behavior data of historical loan users, and the training label of the second prediction model is the credit score set for the dishonest behavior data of historical loan users; the first credit score and the second credit score are summed to obtain the target credit score of the loan user.

[0013] Optionally, the training process of the first prediction model includes the following steps: extracting L model training features from the integrity behavior data of historical loan users, where L is an integer greater than 1; using a preset function to filter at least one target training feature from the L model training features, wherein the preset function is used to determine the feature screening range of the L model training features; iteratively training the neural network according to the at least one target training feature until the neural network enters a convergence state, wherein the difference between the credit score predicted and generated by the neural network in the convergence state based on the at least one target training feature and the credit score set for the integrity behavior data of historical loan users is less than a preset threshold.

[0014] Further, according to another aspect of an embodiment of the present application, a device for determining a credit score of a loan user is provided, comprising: an acquisition unit, for acquiring N types of financial data of the loan user with the authorization of the user, wherein N is an integer greater than 1; a calculation unit, for calculating mathematical statistical data of the i-th type of financial data when missing values ​​and / or abnormal values ​​are detected in the i-th type of financial data among the N types of financial data, wherein the i-th type of financial data is any type of financial data among the N types of financial data; a correction and extraction unit, for correcting the missing values ​​and / or abnormal values ​​in the i-th type of financial data according to the mathematical statistical data, and performing feature extraction on the corrected i-th type of financial data to obtain a set of feature sequences corresponding to the i-th type of financial data; a determination unit, for determining the target credit score of the loan user according to the N groups of feature sequences corresponding to the N types of financial data.

[0015] Optionally, the calculation unit includes: a first calculation subunit and a second calculation subunit. The first calculation subunit is used to calculate the average value of the i-th category of financial data when missing values ​​are detected in the i-th category of financial data; the second calculation subunit is used to calculate the average value and standard deviation of the i-th category of financial data when abnormal values ​​are detected in the i-th category of financial data.

[0016] Optionally, the correction extraction unit includes: a first detection subunit and a second detection subunit. The first detection subunit is used to fill the missing values ​​in the i-th category of financial data with the average value of the i-th category of financial data when missing values ​​are detected in the i-th category of financial data; the second detection subunit is used to calculate the difference between the average value of the i-th category of financial data and the abnormal value when abnormal values ​​are detected in the i-th category of financial data, and perform any of the following correction operations on the abnormal value according to the absolute value of the difference and the standard deviation of the i-th category of financial data: retain the abnormal value, delete the abnormal value, and modify the abnormal value.

[0017] Optionally, the second detection subunit includes: a first calculation module, a first detection module and a second detection module. The first calculation module is used to calculate the product of the standard deviation and a preset coefficient to obtain a first value, wherein the preset coefficient is used to amplify or reduce the standard deviation; the first detection module is used to delete or modify the abnormal value when the absolute value of the difference is detected to be greater than the first value, wherein the absolute value of the difference between the modified abnormal value and the average value of the i-th category of financial data is less than or equal to the first value; the second detection module is used to retain the abnormal value when the absolute value of the difference is detected to be less than or equal to the first value.

[0018] Optionally, the determination unit includes: a first input subunit, a first determination subunit, and a third calculation subunit. The first input subunit is used to input N groups of feature sequences into N model branches respectively, wherein the jth model branch among the N model branches is used to process the jth group of feature sequences among the N groups of feature sequences, and the jth type of financial data corresponding to the jth group of feature sequences is the same type of data as the training data of the jth model branch; the first determination subunit is used to determine an initial credit score of the loan user through each model branch based on a set of feature sequences input, combined with the prior knowledge of the model branch during the training process, wherein the N model branches generate N initial credit scores based on the N groups of feature sequences; the third calculation subunit is used to calculate the average of the N initial credit scores to obtain the target credit score of the loan user.

[0019] Optionally, the determination unit includes: a first classification subunit, a second determination subunit, a third determination subunit and a fourth calculation subunit. The first classification subunit is used to classify N groups of feature sequences into a first type of feature sequence and a second type of feature sequence, wherein the number of features in the first type of feature sequence is positively correlated with the credit score of the loan user, and the number of features in the second type of feature sequence is negatively correlated with the credit score of the loan user; the second determination subunit is used to input the first type of feature sequence into the first prediction model, and determine the first credit score based on the first type of feature sequence through the prior knowledge learned by the first prediction model in the model training phase, wherein the first credit score is a value greater than 0, the training data of the first prediction model is the integrity behavior data of historical loan users, and the first prediction model The training label is a credit score set for the honest behavior data of historical loan users; the third determination subunit is used to input the second type of feature sequence into the second prediction model, and determine the second credit score based on the second type of feature sequence through the prior knowledge learned by the second prediction model in the model training phase, wherein the second credit score is a value less than 0, the training data of the second prediction model is the dishonest behavior data of historical loan users, and the training label of the second prediction model is the credit score set for the dishonest behavior data of historical loan users; the fourth calculation subunit is used to sum the first credit score and the second credit score to obtain the target credit score of the loan user.

[0020] Optionally, the training process of the first prediction model includes the following steps: extracting L model training features from the integrity behavior data of historical loan users, where L is an integer greater than 1; using a preset function to filter at least one target training feature from the L model training features, wherein the preset function is used to determine the feature screening range of the L model training features; iteratively training the neural network according to the at least one target training feature until the neural network enters a convergence state, wherein the difference between the credit score predicted and generated by the neural network in the convergence state based on the at least one target training feature and the credit score set for the integrity behavior data of historical loan users is less than a preset threshold.

[0021] According to another aspect of an embodiment of the present application, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned method for determining the credit score of the loan user.

[0022] According to another aspect of an embodiment of the present application, there is also provided an electronic device, comprising: a memory storing an executable program; and a processor for running the program, wherein the program executes the above-mentioned method for determining the credit score of a loan user when running.

[0023] According to another aspect of an embodiment of the present application, a computer program product is also provided, including computer instructions, which, when executed by a processor, implement the steps of the above-mentioned method for determining the credit score of a loan user.

[0024] In an embodiment of the present application, first, with the user's authorization, N types of financial data of the loan user are obtained, where N is an integer greater than 1; when it is detected that there are missing values ​​and / or abnormal values ​​in the i-th type of financial data among the N types of financial data, the mathematical statistical data of the i-th type of financial data is calculated, where the i-th type of financial data is any type of financial data among the N types of financial data; the missing values ​​and / or abnormal values ​​in the i-th type of financial data are corrected according to the mathematical statistical data, and features are extracted from the corrected i-th type of financial data to obtain a set of feature sequences corresponding to the i-th type of financial data; and the target credit score of the loan user is determined according to the N groups of feature sequences corresponding to the N types of financial data.

[0025] As can be seen from the above, this application first ensures that N types of financial data of loan users are collected in strict compliance with privacy regulations and with explicit authorization from users, where N represents the number of data types and is an integer greater than 1. This step emphasizes the importance of user privacy protection while ensuring the legitimacy and compliance of data sources.

[0026] Next, this application conducts an in-depth quality check on the collected data, automatically identifying and processing missing values ​​and outliers. By calculating the mathematical statistics of each type of financial data (such as mean and standard deviation), a high-quality data foundation is provided for subsequent feature extraction and model training. The automation and intelligence of data preprocessing significantly improves the efficiency and reliability of the evaluation process.

[0027] Secondly, for financial data with missing values ​​and / or outliers detected, this application uses advanced data correction techniques, such as mean value filling of missing values ​​and statistical correction of outliers, to ensure the integrity and accuracy of the data. Subsequently, feature extraction is performed on the corrected data to generate a set of feature sequences closely related to the i-th category of financial data. This process effectively converts the original data into a form that can be understood by the model, providing direct input for the determination of credit scores.

[0028] Finally, this application uses the extracted N groups of feature sequences to determine the target credit score of the loan user through comprehensive analysis and algorithm model. This score is not only based on single-dimensional data, but also comprehensively considers the user's various financial conditions, thereby obtaining a more objective and accurate credit assessment result, providing a basis for decision-making for lending institutions.

[0029] In the embodiments of the present application, the above method is adopted to construct an intelligent framework that protects user privacy and can accurately assess credit status through steps such as authorized data acquisition, data quality assurance, feature extraction and correction, and multi-dimensional feature analysis, thereby achieving the technical effect of optimizing the data processing process and improving the accuracy of determining the user's credit score, thereby solving the technical problem of low accuracy in determining the user's credit score due to too many types of user data in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0031] Figure 1 A hardware structure block diagram of a computer terminal for implementing a method for determining a credit score of a loan user is shown;

[0032] Figure 2 is a flow chart of a method for determining a credit score of a loan user according to an embodiment of the present application;

[0033] Figure 3 This is a brief flowchart of a personal credit assessment based on a Bayesian optimization algorithm SVM algorithm according to an embodiment of the present application;

[0034] Figure 4 It is an overall flow chart based on SVM algorithm and Bayesian algorithm optimization according to an embodiment of the present application;

[0035] Figure 5 is a schematic diagram of an optional device for determining a credit score of a loan user according to an embodiment of the present application;

[0036] Figure 6 It is a structural block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0037] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution 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 part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0038] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0039] According to an embodiment of the present application, a method embodiment of a method for determining a credit score of a loan user is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0040] It should be noted that a credit score determination system can be used as the execution subject of the method for determining the credit score of a loan user in the embodiment of the present application. It is understandable that the method for determining the credit score of a loan user provided in the embodiment of the present application can also be executed by other systems or devices, and the embodiment of the present application does not specifically limit this.

[0041] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 The hardware structure block diagram of a computer terminal (or mobile device) for implementing a method for determining a credit score of a loan user is shown. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more (102a, 102b, ..., 102n are used to illustrate) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0042] It should be noted that the above-mentioned one or more processors 102 and / or other data processing circuits may generally be referred to as "data processing circuits" in this article. The data processing circuit may be embodied in whole or in part as software, hardware, firmware or any other combination. In addition, the data processing circuit may be a single independent processing module, or may be fully or partially integrated into any of the other components in the computer terminal 10 (or mobile device). As in the method for determining the credit score of a loan user involved in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0043] The memory 104 can be used to store software programs and modules of application software, such as program instructions / data storage devices corresponding to the method for determining the credit score of a loan user in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, the method for determining the credit score of a loan user described above is implemented. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0044] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0045] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 (or mobile device).

[0046] Figure 2 is a method for determining the credit score of a loan user according to an embodiment of the present application, such as Figure 2 As shown, the method comprises the following steps:

[0047] Step S201, with the authorization of the user, obtain N types of financial data of the loan user, where N is an integer greater than 1.

[0048] Optionally, with the user's authorization, the customer's personal loan-related data is obtained from the financial institution's internal database.

[0049] Optionally, the N types of financial data of the loan user include but are not limited to characteristics such as age x1, income x2, debt x3, number of overdue payments x4, and credit assessment data y (y=1 indicates good credit, y=-1 indicates poor credit).

[0050] Step S202, when it is detected that there are missing values ​​and / or abnormal values ​​in the i-th category of financial data among the N categories of financial data, the mathematical statistical data of the i-th category of financial data is calculated.

[0051] In step S202, the i-th type of financial data is any type of financial data among N types of financial data.

[0052] Alternatively, in the process of financial data collection and analysis, missing values ​​refer to situations where complete information is not available in data records. This may be due to negligence in the data collection process, technical failures, the implementation of data privacy protection measures or other unforeseen reasons. The existence of missing values ​​will seriously affect the integrity of the data and the accuracy of the model.

[0053] Alternatively, in the process of financial data collection and analysis, outliers refer to values ​​that deviate significantly from other data points in the data set. Such values ​​may be caused by measurement errors, data entry errors, or extreme events. The existence of outliers may also distort the results of data analysis and lead to poor model training results.

[0054] Optionally, in order to ensure the quality of the N types of financial data, the credit score determination system uses data detection technology. Through data integrity checking, it is possible to automatically identify whether there are missing values ​​in the i-th type of financial data. This process usually involves traversing the data fields and checking whether each field contains a non-null value, thereby ensuring that every part of the data is fully considered.

[0055] Optionally, for detection of outliers, the credit score determination system is based on statistical methods to identify values ​​that differ too much from the average value by calculating mathematical statistical data of the i-th category of financial data, such as the average value, standard deviation, etc.

[0056] Optionally, after detecting missing values ​​or outliers, the credit score determination system calculates mathematical statistics of the i-th category of financial data to support subsequent data correction work. The main statistics include: mean value, standard deviation and other statistics.

[0057] Step S203, correcting missing values ​​and / or abnormal values ​​in the i-th category of financial data according to mathematical statistical data, and extracting features from the corrected i-th category of financial data to obtain a set of feature sequences corresponding to the i-th category of financial data.

[0058] Optionally, for the correction of missing values, the credit scoring system adopts a dynamic and refined method to select the most appropriate filling strategy according to the characteristics of different types of financial data. For the basic information data of loan users (such as age, income, etc.), if there are missing values, the credit scoring system first calculates the statistics such as the mean, median or mode of such data, and then selects the most appropriate filling method according to the distribution characteristics of the data. For data with normal distribution, the credit scoring system uses the mean value for filling; for data with non-normal distribution or extreme values, the credit scoring system uses the median filling or interpolation filling based on the relationship between data points, which can more accurately reflect the true distribution of the data, thereby avoiding model bias caused by improper filling of missing values.

[0059] Optionally, when dealing with outliers, the credit scoring system adopts a rigorous processing strategy. First, the relevant data statistics of the outliers are calculated to identify the values ​​that significantly deviate from the central trend of the data set. Then, based on the nature of the outliers and the degree of impact on model training, the credit scoring system dynamically takes actions to retain, delete, or modify them.

[0060] Optionally, feature extraction is a key step in machine learning model training, which aims to select the most influential feature subset for model prediction from the original multi-dimensional financial data. In this application, feature extraction covers not only simple transformation and dimensionality reduction of data, but also feature engineering techniques such as feature construction, feature selection and feature encoding, in order to obtain a set of feature sequences that can clearly reflect the credit status of loan users.

[0061] Step S204, determining the target credit score of the loan user according to the N groups of feature sequences corresponding to the N types of financial data.

[0062] Optionally, determining the target credit score of the loan user starts with the construction of the feature vector, and goes through key steps such as model training and parameter optimization, model verification and testing, and finally generates a comprehensive and accurate credit score.

[0063] Optionally, the present application uses the support vector machine (SVM) algorithm as the core model and combines it with Bayesian optimization to intelligently adjust the model parameters. The model will predict the credit status of the user based on his / her characteristics and convert it into a quantitative credit score.

[0064] Through the above steps, in an embodiment of the present application, first, with the authorization of the user, N types of financial data of the loan user are obtained, where N is an integer greater than 1; when it is detected that there are missing values ​​and / or abnormal values ​​in the i-th type of financial data among the N types of financial data, the mathematical statistical data of the i-th type of financial data is calculated, where the i-th type of financial data is any type of financial data among the N types of financial data; the missing values ​​and / or abnormal values ​​in the i-th type of financial data are corrected according to the mathematical statistical data, and features are extracted from the corrected i-th type of financial data to obtain a set of feature sequences corresponding to the i-th type of financial data; and the target credit score of the loan user is determined according to the N groups of feature sequences corresponding to the N types of financial data.

[0065] As can be seen from the above, this application first ensures that N types of financial data of loan users are collected in strict compliance with privacy regulations and with explicit authorization from users, where N represents the number of data types and is an integer greater than 1. This step emphasizes the importance of user privacy protection while ensuring the legitimacy and compliance of data sources.

[0066] Next, this application conducts an in-depth quality check on the collected data, automatically identifying and processing missing values ​​and outliers. By calculating the mathematical statistics of each type of financial data (such as mean and standard deviation), a high-quality data foundation is provided for subsequent feature extraction and model training. The automation and intelligence of data preprocessing significantly improves the efficiency and reliability of the evaluation process.

[0067] Secondly, for financial data with missing values ​​and / or outliers detected, this application uses advanced data correction techniques, such as mean value filling of missing values ​​and statistical correction of outliers, to ensure the integrity and accuracy of the data. Subsequently, feature extraction is performed on the corrected data to generate a set of feature sequences closely related to the i-th category of financial data. This process effectively converts the original data into a form that can be understood by the model, providing direct input for the determination of credit scores.

[0068] Finally, this application uses the extracted N groups of feature sequences to determine the target credit score of the loan user through comprehensive analysis and algorithm model. This score is not only based on single-dimensional data, but also comprehensively considers the user's various financial conditions, thereby obtaining a more objective and accurate credit assessment result, providing a basis for decision-making for lending institutions.

[0069] In the embodiments of the present application, the above method is adopted to construct an intelligent framework that protects user privacy and can accurately assess credit status through steps such as authorized data acquisition, data quality assurance, feature extraction and correction, and multi-dimensional feature analysis, thereby achieving the technical effect of optimizing the data processing process and improving the accuracy of determining the user's credit score, thereby solving the technical problem of low accuracy in determining the user's credit score due to too many types of user data in the prior art.

[0070] In an optional embodiment, the credit score determination system first calculates the average value of the i-th category of financial data when missing values ​​are detected in the i-th category of financial data, and secondly, and / or, calculates the average value and standard deviation of the i-th category of financial data when outliers are detected in the i-th category of financial data.

[0071] Optionally, the collected data can be preprocessed by cleaning, standardization, missing value filling, etc. to improve data quality.

[0072] Alternatively, the average (also called the arithmetic mean) is the sum of all the values ​​in a data set divided by the number of data points. It is one of the most basic statistics for describing trends in a data set.

[0073] Optionally, when the credit score determination system detects that there are missing values ​​in the i-th category of financial data, such as x2 (income), the data vector x i =(x i1 , x i2 , x i3 , x i4 ), if mean filling is used, let the set of all income values ​​be (n is the number of data points), then the fill value is (i.e. the average value of the i-th category of financial data).

[0074] Alternatively, standard deviation is an important statistic to measure the degree of dispersion of data distribution. The larger the standard deviation, the greater the difference between data points and the more dispersed the data distribution; conversely, the more concentrated the data points, the more uniform the distribution. In financial data processing, standard deviation is often used to detect outliers and help evaluate the stability of data and the reliability of prediction models.

[0075] Optionally, when the credit score determination system detects an outlier in the i-th category of financial data, it calculates the average value and standard deviation of the i-th category of financial data. The average value can be Calculate the standard deviation using calculate.

[0076] From the above content, we can see that for the missing value and outlier processing method, by calculating the mean of missing values, as well as the mean and standard deviation of outliers, the quality of data can be improved, providing a solid foundation for subsequent feature extraction, model training and credit score determination.

[0077] In an optional embodiment, the credit score determination system first fills the missing values ​​in the i-th category of financial data with the average value of the i-th category of financial data when missing values ​​are detected in the i-th category of financial data, and secondly, and / or, when an outlier is detected in the i-th category of financial data, calculates the difference between the average value of the i-th category of financial data and the outlier, and performs any of the following correction operations on the outlier based on the absolute value of the difference and the standard deviation of the i-th category of financial data: retain the outlier, delete the outlier, and modify the outlier.

[0078] Optionally, when the credit score determination system detects that there are missing values ​​in the i-th category of financial data, such as x2 (income), the data vector x i =(x i1 , x i2 , x i3 , x i4 ), if mean filling is used, let the set of all income values ​​be (n is the number of data points), then the fill value is (i.e. the average value of the i-th category of financial data).

[0079] Optionally, when the credit score determination system detects that there is an outlier in the i-th category of financial data, it calculates the difference between the average value of the i-th category of financial data and the outlier, that is, if the outlier x 2k It obeys normal distribution, assuming the mean is μ and the standard deviation is σ, then for the data point x 2k , calculate |x 2k -μ|.

[0080] Optionally, according to the absolute value of the difference between the mean value and the outlier of the i-th category of financial data and the standard deviation of the i-th category of financial data, that is, according to |x 2k -μ| and σ, and perform any of the following correction operations on the outliers: retain the outliers, delete the outliers, or modify the outliers.

[0081] From the above content, we can see that for missing values, the credit scoring system adopts a feature-based average filling strategy. This method makes full use of the existing information in the data set and fills the missing values ​​by calculating the average value of non-missing data, effectively avoiding the prediction bias caused by the lack of data integrity and ensuring that the input data for model training is complete and representative. For outliers, intelligent operations are performed based on the absolute value of the difference between the average value of the i-th category of financial data and the outlier and the standard deviation of this type of data. It not only eliminates the impact of outliers, but also ensures the quality and consistency of the data, providing a high-quality data foundation for subsequent feature extraction, model training and credit score determination.

[0082] In an optional embodiment, the credit score determination system first calculates the product of the standard deviation and a preset coefficient to obtain a first value, wherein the preset coefficient is used to amplify or reduce the standard deviation, and then deletes or modifies the outlier when it is detected that the absolute value of the difference is greater than the first value, wherein the absolute value of the difference between the modified outlier and the average value of the i-th category of financial data is less than or equal to the first value, and finally retains the outlier when it is detected that the absolute value of the difference is less than or equal to the first value.

[0083] Optionally, the preset coefficient is an artificially defined value used to amplify or reduce the standard deviation.

[0084] Optionally, the first value refers to an abnormality index calculated based on the absolute value of the difference between the abnormal value and the average value of the category data to which it belongs, and the standard deviation of the category data. The calculation of the first value not only quantifies the degree of deviation of the abnormal value, but also takes into account the inherent volatility in the data set, providing an objective basis for determining the processing strategy of the abnormal value.

[0085] Optionally, for outliers whose first values ​​far exceed the preset threshold, the credit score determination system adopts a strategy of direct deletion. Because such outliers often represent data entry errors, measurement errors, or atypical events, the existence of such outliers may seriously distort the training results of the model, causing the model to make inaccurate predictions when faced with normal data. By deleting these significantly abnormal values, the purity and representativeness of the model training data can be guaranteed, avoiding the adverse effects of outliers on model training.

[0086] Optionally, for outliers whose first values ​​are within a certain range, the credit score determination system adopts a modification strategy. Specifically, based on the difference between the outlier and the average value of the data in the category to which it belongs, combined with the standard deviation of the data of this category, the credit score determination system implements a value adjustment within a reasonable range. This strategy aims to retain the valuable information that may exist in the outliers while reducing their negative impact on the performance of the model. Through intelligent correction rather than simple elimination, the diversity of data is retained and the model's adaptability to complex financial environments is enhanced.

[0087] From the above content, it can be seen that the credit score determination system uses a strategy to delete and modify outliers based on the absolute value of the difference between the outliers and the average value of the category data to which they belong, as well as the degree of anomaly index calculated from the standard deviation of the data of this category, which can improve the quality of data and enhance the generalization ability of the model.

[0088] In an optional embodiment, the credit score determination system first inputs N groups of feature sequences into N model branches respectively, wherein the jth model branch among the N model branches is used to process the jth group of feature sequences among the N groups of feature sequences, and the jth type of financial data corresponding to the jth group of feature sequences is the same type of data as the training data of the jth model branch, and then each model branch determines an initial credit score for the loan user based on a set of input feature sequences combined with the prior knowledge of the model branch during the training process, wherein the N model branches generate N initial credit scores based on the N groups of feature sequences, and finally the average of the N initial credit scores is calculated to obtain the target credit score of the loan user.

[0089] Optionally, the credit score determination system adopts a multi-model branch fusion method, where N groups of carefully selected feature sequences are input into N independent model branches for evaluation. Each group of feature sequences focuses on different aspects of personal loan credit assessment, such as income stability, debt ratio, credit history, asset status, etc., to build a comprehensive and in-depth user credit profile.

[0090] Optionally, in each model branch, the credit score determination system uses the corresponding feature sequence to perform a preliminary credit score on the loan user, a step that aims to independently evaluate the user's credit status from multiple perspectives. Each initial credit score is based on the prediction results of a specific branch model, reflecting the user's credit status under that feature sequence.

[0091] Optionally, the credit score determination system calculates an average of N initial credit scores as a preliminary result of the user's credit score, integrates the score information from the N model branches, and balances the evaluation results of different model branches.

[0092] From the above content, we can see that the credit score determination system inputs N groups of feature sequences into N model branches respectively, which can deeply explore the impact of each group of features on the credit score and avoid mutual interference between features. Then, the evaluation results of different model branches are balanced through averaging operations, avoiding the bias or error that may be caused by a single model, thereby providing users with a more comprehensive and objective credit score.

[0093] In an optional embodiment, the credit score determination system first classifies N groups of feature sequences into a first feature sequence and a second feature sequence, wherein the number of features in the first feature sequence is positively correlated with the credit score of the loan user, and the number of features in the second feature sequence is negatively correlated with the credit score of the loan user. Then, the first feature sequence is input into the first prediction model, and the first credit score is determined based on the first feature sequence by using the prior knowledge learned by the first prediction model in the model training phase, wherein the first credit score is a value greater than 0, the training data of the first prediction model is the integrity behavior data of the historical loan users, and the training label of the first prediction model is the credit score set for the integrity behavior data of the historical loan users. Then, the second feature sequence is input into the second prediction model, and the second credit score is determined based on the second feature sequence by using the prior knowledge learned by the second prediction model in the model training phase, wherein the second credit score is a value less than 0, the training data of the second prediction model is the dishonest behavior data of the historical loan users, and the training label of the second prediction model is the credit score set for the dishonest behavior data of the historical loan users. Finally, the first credit score and the second credit score are summed to obtain the target credit score of the loan user.

[0094] Optionally, the credit score determination system subdivides the N groups of feature sequences into first-category feature sequences and second-category feature sequences according to their different contributions and relevances to credit assessment.

[0095] Optionally, the number of features in the first type of feature sequence is positively correlated with the credit score of the loan user; and the number of features in the second type of feature sequence is negatively correlated with the credit score of the loan user.

[0096] Optionally, the first prediction model focuses on processing features that directly reflect the credit level, and can quickly and accurately give evaluation results based on core credit indicators; while the second prediction model is more focused on evaluating the impact of background information on credit, and further improves the user's credit profile by introducing indirect features.

[0097] Optionally, the credit score determination system calculates the sum of the outputs of the first prediction model and the second prediction model, i.e., the first credit score and the second credit score, as a reference for the final credit score of the loan user. This calculation step reflects the concept of the application of multi-level and multi-angle evaluation of the user's credit status. Through the summation operation, the influence of direct and indirect credit factors can be combined to obtain a more comprehensive and balanced credit score result.

[0098] From the above content, it can be seen that the credit score determination system divides N groups of feature sequences into the first type of feature sequences and the second type of feature sequences according to their different contributions and relevance to credit assessment, which helps to deeply explore the independent impact of different feature sequences on credit scores. Then the first prediction model and the second prediction model are used for evaluation, and the first credit score and the second credit score are summed and calculated, thus constructing a hierarchical, comprehensive and accurate credit assessment system, which significantly improves the decision-making efficiency and accuracy of banks in personal loan credit risk management.

[0099] In an optional embodiment, the credit score determination system first extracts L model training features from the integrity behavior data of historical loan users, where L is an integer greater than 1, and then uses a preset function to filter out at least one target training feature from the L model training features, wherein the preset function is used to determine the feature screening range of the L model training features, and finally iteratively trains the neural network according to the at least one target training feature until the neural network enters a convergence state, wherein the difference between the credit score predicted and generated by the neural network in the convergence state based on the at least one target training feature and the credit score set for the integrity behavior data of historical loan users is less than a preset threshold.

[0100] Optionally, the data set used for model training needs to be divided. Set the data set to D = {(x1, y1), (x2, y2), ..., (x n ,y n )} Divide it into training set D train , validation set D val and the test set D test Satisfaction | D train |=0.6n,|D val |=0.2n,|D test |=0.2n.

[0101] Optionally, the credit score determination system extracts L model training features from the integrity behavior data of historical loan users. These features cover multi-dimensional information such as the borrower's financial status, repayment history, credit record, etc., aiming to construct a feature space that comprehensively reflects personal credit risk.

[0102] Optionally, the preset function selects at least one target training feature from the L model training features. If the radial basis kernel (RBF) function is selected as the preset function, its formula is K(x i , x j )=exp(-γ||x i -x j || 2 ), where γ is the kernel parameter (where x i and x j represent two different training features respectively).

[0103] Optionally, the credit score determination system needs to set an initial parameter range, mainly including the initial search range of the penalty parameter C as [C min , C max ], the initial search range of the kernel parameter γ is set to [γ min , γ max ].

[0104] Optionally, a maximum number of iterations is set. When the number of iterations reaches the maximum number, the algorithm iteration is stopped, and the optimal sampling point and the corresponding optimal value of the objective function, that is, the optimal parameters C and γ of the SVM model, are output.

[0105] Optionally, for the prediction effect of the model, the f1 evaluation index is introduced as the evaluation standard of the model and the AUC value is used to evaluate the overall prediction effect of the model.

[0106] The calculation formula of the f1 evaluation index is

[0107]

[0108] Among them, recall and precision are called recall rate and precision rate respectively.

[0109]

[0110] Among them, TP is the number of positive classes predicted as positive classes; FN is the number of positive classes predicted as negative classes; FP is the number of negative classes predicted as positive classes. The higher the recall, the stronger the model's ability to identify positive samples; the higher the precision, the stronger the model's ability to distinguish negative samples. f1 is a combination of the two. The higher f1, the more robust the established classification model. If either recall or precision decreases, the value of f1 will decrease.

[0111] The AUC value is calculated as follows:

[0112]

[0113] Among them, the AUC value (Area Under Curve) is the area enclosed by the ROC curve (The Receiver Operating Characteristic) and the coordinates. The horizontal axis is FPR (false positive rate), which indicates the probability of predicting a positive value when the actual value is negative, and the vertical axis is TPR (true positive rate), which indicates the probability of predicting a positive value when the actual value is positive.

[0114] From the above content, it can be seen that the credit scoring determination system builds an efficient and accurate personal loan credit assessment system by intelligently extracting and screening key features from historical loan users' integrity behavior data, combined with iterative training of deep learning neural networks, which improves the predictive performance of the model and provides more reliable data support for the bank's personal loan decisions.

[0115] In an optional embodiment, Figure 3 A personal credit assessment method based on the Bayesian optimization algorithm SVM algorithm according to an embodiment of the present application is shown. Figure 3 As shown in the figure, firstly, data collection and data preprocessing are carried out to obtain customer personal loan related data from the internal database, and then the collected data is cleaned, standardized, missing value filling and other preprocessing operations are carried out to improve data quality. Then the SVM model is trained to select appropriate kernel functions and parameters. Then the Bayesian model is trained and the Gaussian process proxy model is updated for optimization. Then it is iterated to output the optimal parameters C and γ of the SVM model. Finally, the credit assessment standard is set to obtain the credit score.

[0116] In an optional embodiment, Figure 4 The overall process based on SVM algorithm and Bayesian algorithm optimization according to an embodiment of the present application is shown as follows: Figure 4 As shown:

[0117] 1. Collect multi-dimensional information of customers, including basic information, financial information, credit records, behavioral habits, etc.;

[0118] 2. Perform pre-processing operations such as cleaning, standardization, and missing value filling on the collected data to improve data quality;

[0119] 3. Select key features related to credit risk from the preprocessed data.

[0120] 4. Map the data from the original feature space to the high-dimensional feature space through the kernel function, make the linear inner product operation nonlinear, and then establish the optimal hyperplane that maximizes the classification interval in the high-dimensional feature space.

[0121] 5. Find the parameters w and b that maximize γ.

[0122] 6. Randomly select n0 sampling points in the set search range of C and γ of the SVM model, take the average test accuracy of ten-fold cross validation as the objective function f, and use different parameter combinations of the model as independent variables x to form the proxy model framework, and obtain the initial distribution of the objective function and the collection point set D1.

[0123] 7. Select the next sampling point x by maximizing the acquisition function t , and obtain the sampling point function f(x t ).

[0124] 8. Set the new sampling point [x t ,f(x t )] is added to the sampling point set D1 and the Gaussian process proxy model is updated so that the proxy model is more consistent with the distribution of the objective function.

[0125] 9. When the number of iterations reaches the maximum, stop the algorithm iteration and output the optimal sampling point and the corresponding optimal value of the objective function.

[0126] 10. Introduce the f1 evaluation index as the evaluation criterion of the model.

[0127] 11. Introduce the AUC value to evaluate the overall prediction effect of the model.

[0128] According to another aspect of the embodiment of the present application, a device for determining a credit score of a loan user is also provided, wherein: Figure 5 is a schematic diagram of an optional device for determining a credit score of a loan user according to an embodiment of the present application, such as Figure 5 As shown, the device for determining the credit score of a loan user includes: an acquisition unit 501, a calculation unit 502, a correction extraction unit 503, and a determination unit 504.

[0129] The acquisition unit 501 is used to acquire N types of financial data of the loan user with the authorization of the user, wherein N is an integer greater than 1; the calculation unit 502 is used to calculate the mathematical statistical data of the i-th type of financial data when it is detected that there are missing values ​​and / or abnormal values ​​in the i-th type of financial data among the N types of financial data, wherein the i-th type of financial data is any type of financial data among the N types of financial data; the correction and extraction unit 503 is used to correct the missing values ​​and / or abnormal values ​​in the i-th type of financial data according to the mathematical statistical data, and perform feature extraction on the corrected i-th type of financial data to obtain a set of feature sequences corresponding to the i-th type of financial data; the determination unit 504 is used to determine the target credit score of the loan user according to the N groups of feature sequences corresponding to the N types of financial data.

[0130] Optionally, the calculation unit 501 includes: a first calculation subunit and a second calculation subunit. The first calculation subunit is used to calculate the average value of the i-th category of financial data when missing values ​​are detected in the i-th category of financial data; the second calculation subunit is used to calculate the average value and standard deviation of the i-th category of financial data when abnormal values ​​are detected in the i-th category of financial data.

[0131] Optionally, the correction extraction unit 503 includes: a first detection subunit and a second detection subunit. The first detection subunit is used to fill the missing values ​​in the i-th category of financial data with the average value of the i-th category of financial data when missing values ​​are detected in the i-th category of financial data; the second detection subunit is used to calculate the difference between the average value of the i-th category of financial data and the abnormal value when abnormal values ​​are detected in the i-th category of financial data, and perform any of the following correction operations on the abnormal value according to the absolute value of the difference and the standard deviation of the i-th category of financial data: retain the abnormal value, delete the abnormal value, and modify the abnormal value.

[0132] Optionally, the second detection subunit includes: a first calculation module, a first detection module and a second detection module. The first calculation module is used to calculate the product of the standard deviation and a preset coefficient to obtain a first value, wherein the preset coefficient is used to amplify or reduce the standard deviation; the first detection module is used to delete or modify the abnormal value when the absolute value of the difference is detected to be greater than the first value, wherein the absolute value of the difference between the modified abnormal value and the average value of the i-th category of financial data is less than or equal to the first value; the second detection module is used to retain the abnormal value when the absolute value of the difference is detected to be less than or equal to the first value.

[0133] Optionally, the determination unit 504 includes: a first input subunit, a first determination subunit, and a third calculation subunit. The first input subunit is used to input N groups of feature sequences into N model branches respectively, wherein the j-th model branch among the N model branches is used to process the j-th group of feature sequences among the N groups of feature sequences, and the j-th type of financial data corresponding to the j-th group of feature sequences is the same type of data as the training data of the j-th model branch; the first determination subunit is used to determine an initial credit score of the loan user through each model branch based on a group of feature sequences input, combined with the prior knowledge of the model branch during the training process, wherein the N model branches generate N initial credit scores based on the N groups of feature sequences; the third calculation subunit is used to calculate the average of the N initial credit scores to obtain the target credit score of the loan user.

[0134] Optionally, the determination unit 504 includes: a first classification subunit, a second determination subunit, a third determination subunit, and a fourth calculation subunit. The first classification subunit is used to classify N groups of feature sequences into a first type of feature sequence and a second type of feature sequence, wherein the number of features in the first type of feature sequence is positively correlated with the credit score of the loan user, and the number of features in the second type of feature sequence is negatively correlated with the credit score of the loan user; the second determination subunit is used to input the first type of feature sequence into the first prediction model, and determine the first credit score based on the first type of feature sequence through the prior knowledge learned by the first prediction model in the model training stage, wherein the first credit score is a value greater than 0, the training data of the first prediction model is the integrity behavior data of the historical loan users, and the first prediction model The training label is a credit score set for the honest behavior data of historical loan users; the third determination subunit is used to input the second type of feature sequence into the second prediction model, and determine the second credit score based on the second type of feature sequence through the prior knowledge learned by the second prediction model in the model training phase, wherein the second credit score is a value less than 0, the training data of the second prediction model is the dishonest behavior data of historical loan users, and the training label of the second prediction model is the credit score set for the dishonest behavior data of historical loan users; the fourth calculation subunit is used to sum the first credit score and the second credit score to obtain the target credit score of the loan user.

[0135] Optionally, the training process of the first prediction model includes the following steps: extracting L model training features from the integrity behavior data of historical loan users, where L is an integer greater than 1; using a preset function to filter at least one target training feature from the L model training features, wherein the preset function is used to determine the feature screening range of the L model training features; iteratively training the neural network according to the at least one target training feature until the neural network enters a convergence state, wherein the difference between the credit score predicted and generated by the neural network in the convergence state based on the at least one target training feature and the credit score set for the integrity behavior data of historical loan users is less than a preset threshold.

[0136] It should be noted here that the above-mentioned modules or units can be hardware components or software components stored in a memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n), and the above-mentioned modules can also be run as part of the device in the computer terminal 10 provided in the embodiment.

[0137] An embodiment of the present application may provide an electronic device, Figure 6 is a structural block diagram of an electronic device according to an embodiment of the present application. Figure 6As shown, the electronic device may include: one or more ( Figure 6 Only one is shown) a processor, a memory, a storage controller, and a peripheral interface, wherein the peripheral interface is connected to a radio frequency module, an audio module, and a display.

[0138] Among them, the memory can be used to store software programs and modules, such as program instructions / modules corresponding to the methods and devices in the embodiments of the present application, and the processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the above-mentioned method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0139] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: with the user's authorization, obtain N types of financial data of the loan user, where N is an integer greater than 1; when it is detected that there are missing values ​​and / or abnormal values ​​in the i-th type of financial data among the N types of financial data, calculate the mathematical statistical data of the i-th type of financial data, where the i-th type of financial data is any type of financial data among the N types of financial data; correct the missing values ​​and / or abnormal values ​​in the i-th type of financial data according to the mathematical statistical data, and extract features from the corrected i-th type of financial data to obtain a set of feature sequences corresponding to the i-th type of financial data; determine the target credit score of the loan user according to the N groups of feature sequences corresponding to the N types of financial data.

[0140] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: when missing values ​​are detected in the i-th category of financial data, calculate the average value of the i-th category of financial data; and / or; when outliers are detected in the i-th category of financial data, calculate the average value and standard deviation of the i-th category of financial data.

[0141] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: when missing values ​​are detected in the i-th category of financial data, use the average value of the i-th category of financial data to fill the missing values ​​in the i-th category of financial data; and / or; when an outlier is detected in the i-th category of financial data, calculate the difference between the average value of the i-th category of financial data and the outlier, and perform any of the following correction operations on the outlier based on the absolute value of the difference and the standard deviation of the i-th category of financial data: retain the outlier, delete the outlier, and modify the outlier.

[0142] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: calculate the product of the standard deviation and the preset coefficient to obtain a first value, wherein the preset coefficient is used to amplify the standard deviation or reduce the standard deviation; when it is detected that the absolute value of the difference is greater than the first value, delete the outlier or modify the outlier, wherein the absolute value of the difference between the modified outlier and the average value of the i-th category of financial data is less than or equal to the first value; when it is detected that the absolute value of the difference is less than or equal to the first value, retain the outlier.

[0143] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: input N groups of feature sequences into N model branches respectively, wherein the j-th model branch among the N model branches is used to process the j-th group of feature sequences among the N groups of feature sequences, and the j-th type of financial data corresponding to the j-th group of feature sequences is the same type of data as the training data of the j-th model branch; determine an initial credit score of the loan user through each model branch based on a set of input feature sequences combined with the prior knowledge of the model branch during the training process, wherein the N model branches generate N initial credit scores based on the N groups of feature sequences; calculate the average of the N initial credit scores to obtain the target credit score of the loan user.

[0144] The processor can call the information and application program stored in the memory through the transmission device to perform the following steps: classify N groups of feature sequences into a first type of feature sequence and a second type of feature sequence, wherein the number of features in the first type of feature sequence is positively correlated with the credit score of the loan user, and the number of features in the second type of feature sequence is negatively correlated with the credit score of the loan user; input the first type of feature sequence into the first prediction model, and determine the first credit score based on the first type of feature sequence through the prior knowledge learned by the first prediction model in the model training stage, wherein the first credit score is a value greater than 0, and the training data of the first prediction model is historical loan usage. The first prediction model is based on the credit behavior data of historical loan users, and the training label of the first prediction model is the credit score set for the credit behavior data of historical loan users; the second feature sequence is input into the second prediction model, and the second credit score is determined according to the second feature sequence through the prior knowledge learned by the second prediction model in the model training stage, wherein the second credit score is a value less than 0, the training data of the second prediction model is the dishonest behavior data of historical loan users, and the training label of the second prediction model is the credit score set for the dishonest behavior data of historical loan users; the first credit score and the second credit score are summed to obtain the target credit score of the loan user.

[0145] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: extract L model training features from the integrity behavior data of historical loan users, where L is an integer greater than 1; use a preset function to filter at least one target training feature from the L model training features, where the preset function is used to determine the feature screening range of the L model training features; iteratively train the neural network according to the at least one target training feature until the neural network enters a convergence state, where the difference between the credit score predicted and generated by the neural network in the convergence state based on the at least one target training feature and the credit score set for the integrity behavior data of historical loan users is less than a preset threshold.

[0146] In the embodiment of the present application, firstly, it is ensured that N types of financial data of loan users are collected under strict compliance with privacy regulations and with explicit authorization from users, where N represents the number of data types and is an integer greater than 1. This step emphasizes the importance of user privacy protection while ensuring the legitimacy and compliance of the data source.

[0147] Next, this application conducts an in-depth quality check on the collected data, automatically identifying and processing missing values ​​and outliers. By calculating the mathematical statistics of each type of financial data (such as mean and standard deviation), a high-quality data foundation is provided for subsequent feature extraction and model training. The automation and intelligence of data preprocessing significantly improves the efficiency and reliability of the evaluation process.

[0148] Secondly, for financial data with missing values ​​and / or outliers detected, this application uses advanced data correction techniques, such as mean value filling of missing values ​​and statistical correction of outliers, to ensure the integrity and accuracy of the data. Subsequently, feature extraction is performed on the corrected data to generate a set of feature sequences closely related to the i-th category of financial data. This process effectively converts the original data into a form that can be understood by the model, providing direct input for the determination of credit scores.

[0149] Finally, this application uses the extracted N groups of feature sequences to determine the target credit score of the loan user through comprehensive analysis and algorithm model. This score is not only based on single-dimensional data, but also comprehensively considers the user's various financial conditions, thereby obtaining a more objective and accurate credit assessment result, providing a basis for decision-making for lending institutions.

[0150] In the embodiments of the present application, the above method is adopted to construct an intelligent framework that protects user privacy and can accurately assess credit status through steps such as authorized data acquisition, data quality assurance, feature extraction and correction, and multi-dimensional feature analysis, thereby achieving the technical effect of optimizing the data processing process and improving the accuracy of determining the user's credit score, thereby solving the technical problem of low accuracy in determining the user's credit score due to too many types of user data in the prior art.

[0151] It can be understood by those skilled in the art that Figure 6 The structure shown is for illustration only, and the electronic device may also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile Internet device (Mobile Internet Devices, MID), a PAD, or other terminal devices. Figure 6 The structure of the electronic device is not limited. Figure 6 More or fewer components (such as network interfaces, display devices, etc.) shown in, or having Figure 6 Different configurations shown.

[0152] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0153] According to another aspect of the present application, a computer-readable storage medium is also provided, wherein a computer program is stored in the computer-readable storage medium, wherein when the computer program is executed, the device where the computer-readable storage medium is located executes the above-mentioned method for determining the credit score of the loan user.

[0154] According to another aspect of the present application, a computer program product is also provided, wherein the computer program product includes computer instructions, wherein when the computer instructions are executed, the device where the computer program product is located executes the above-mentioned method for determining the credit score of the loan user.

[0155] The above-mentioned embodiments or examples disclosed in the present application are not exhaustive, but are only illustrative of some embodiments or examples, and are not intended to be specific limitations on the scope of protection disclosed in the present application. In the absence of contradiction, each step in a certain embodiment or example in the present application can be implemented as an independent example, and the steps can be combined arbitrarily. For example, the scheme after removing some steps in a certain embodiment or example can also be implemented as an independent example, and the order of the steps in a certain embodiment or example can be arbitrarily exchanged. In addition, the optional methods or optional examples in a certain embodiment or example can be combined arbitrarily; in addition, the various embodiments or examples can be combined arbitrarily, for example, some or all steps of different embodiments or examples can be combined arbitrarily, and a certain embodiment or example can be combined arbitrarily with the optional methods or optional examples of other embodiments or examples.

[0156] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0157] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0158] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0159] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed over multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0160] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0161] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or all or part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of each embodiment method of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc. Various media that can store program codes.

[0162] The above are only preferred implementations of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for determining a credit score of a loan user, characterized in that: include: With the authorization of the user, obtain N types of financial data of the loan user, where N is an integer greater than 1; When it is detected that there are missing values ​​and / or abnormal values ​​in the i-th category of financial data among the N categories of financial data, calculating mathematical statistics of the i-th category of financial data, wherein the i-th category of financial data is any category of financial data among the N categories of financial data; Correcting missing values ​​and / or abnormal values ​​in the i-th category of financial data according to the mathematical statistical data, and performing feature extraction on the corrected i-th category of financial data to obtain a set of feature sequences corresponding to the i-th category of financial data; The target credit score of the loan user is determined according to the N groups of feature sequences corresponding to the N types of financial data.

2. The method for determining the credit score of a loan user according to claim 1, characterized in that: When it is detected that there are missing values ​​and / or abnormal values ​​in the i-th category of financial data among the N categories of financial data, calculating mathematical statistics of the i-th category of financial data comprises: When it is detected that the missing value exists in the i-th category of financial data, calculating the average value of the i-th category of financial data; and / or; When it is detected that the abnormal value exists in the i-th category financial data, the average value and standard deviation of the i-th category financial data are calculated.

3. The method for determining the credit score of a loan user according to claim 1, characterized in that: Correcting missing values ​​and / or outliers in the i-th category of financial data according to the mathematical statistical data includes: In the case where missing values ​​in the i-th category of financial data are detected, filling the missing values ​​in the i-th category of financial data with an average value of the i-th category of financial data; and / or; When an outlier is detected in the i-th category of financial data, a difference between an average value of the i-th category of financial data and the outlier is calculated, and any one of the following correction operations is performed on the outlier according to the absolute value of the difference and the standard deviation of the i-th category of financial data: Keep the abnormal value, delete the abnormal value, or modify the abnormal value.

4. The method for determining the credit score of a loan user according to claim 3, characterized in that: Modifying or deleting the abnormal value according to the absolute value of the difference and the standard deviation of the i-th category of financial data includes: Calculating the product of the standard deviation and a preset coefficient to obtain a first value, wherein the preset coefficient is used to amplify the standard deviation or reduce the standard deviation; In the case where it is detected that the absolute value of the difference is greater than the first value, deleting the abnormal value or modifying the abnormal value, wherein the absolute value of the difference between the abnormal value after modification and the average value of the i-th category of financial data is less than or equal to the first value; In the case where it is detected that the absolute value of the difference is less than or equal to the first value, the abnormal value is retained.

5. The method for determining the credit score of a loan user according to claim 1, characterized in that: Determining the credit score of the loan user according to the N groups of feature sequences corresponding to the N types of financial data includes: Inputting N groups of feature sequences into N model branches respectively, wherein the j-th model branch among the N model branches is used to process the j-th group of feature sequences among the N groups of feature sequences, and the j-th type of financial data corresponding to the j-th group of feature sequences is the same type of data as the training data of the j-th model branch; Determine an initial credit score of the loan user by each model branch according to a set of feature sequences inputted and in combination with the prior knowledge of the model branch during the training process, wherein the N model branches generate N initial credit scores according to the N sets of feature sequences; An average value of the N initial credit scores is calculated to obtain a target credit score for the loan user.

6. The method for determining the credit score of a loan user according to claim 1, characterized in that: Determining the credit score of the loan user according to the N groups of feature sequences corresponding to the N types of financial data includes: Classifying the N groups of feature sequences into a first type of feature sequence and a second type of feature sequence, wherein the number of features in the first type of feature sequence is positively correlated with the credit score of the loan user, and the number of features in the second type of feature sequence is negatively correlated with the credit score of the loan user; Inputting the first feature sequence into the first prediction model, determining a first credit score based on the first feature sequence through the prior knowledge learned by the first prediction model in the model training phase, wherein the first credit score is a value greater than 0, the training data of the first prediction model is the integrity behavior data of historical loan users, and the training label of the first prediction model is the credit score set for the integrity behavior data of the historical loan users; Inputting the second feature sequence into the second prediction model, determining the second credit score based on the second feature sequence through the prior knowledge learned by the second prediction model in the model training phase, wherein the second credit score is a value less than 0, the training data of the second prediction model is the dishonest behavior data of the historical loan users, and the training label of the second prediction model is the credit score set for the dishonest behavior data of the historical loan users; The first credit score and the second credit score are summed to obtain a target credit score for the loan user.

7. The method for determining the credit score of a loan user according to claim 6, characterized in that: The training process of the first prediction model includes the following steps: Extracting L model training features from the credit behavior data of the historical loan users, where L is an integer greater than 1; Using a preset function to screen out at least one target training feature from the L model training features, wherein the preset function is used to determine a feature screening range of the L model training features; The neural network is iteratively trained according to the at least one target training feature until the neural network enters a convergence state, wherein the difference between the credit score predicted and generated by the neural network in the convergence state based on the at least one target training feature and the credit score set for the integrity behavior data of the historical loan user is less than a preset threshold.

8. A device for determining a credit score of a loan user, characterized in that: include: An acquisition unit, used to acquire N types of financial data of a loan user with authorization from the user, where N is an integer greater than 1; a calculation unit, configured to calculate mathematical statistics of the i-th category of financial data when missing values ​​and / or abnormal values ​​are detected in the i-th category of financial data among the N categories of financial data, wherein the i-th category of financial data is any category of financial data among the N categories of financial data; a feature extraction unit, configured to correct missing values ​​and / or abnormal values ​​in the i-th category of financial data according to the mathematical statistical data, and extract features from the corrected i-th category of financial data to obtain a set of feature sequences corresponding to the i-th category of financial data; A score determination unit is used to determine the target credit score of the loan user according to the N groups of feature sequences corresponding to the N types of financial data.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored executable program, wherein when the executable program is run, the device where the computer-readable storage medium is located is controlled to execute the method for determining the credit score of a loan user as described in any one of claims 1 to 7.

10. An electronic device, characterized in that: include: A memory storing an executable program; A processor, configured to run the program, wherein the program, when running, executes the method for determining the credit score of a loan user as described in any one of claims 1 to 7.

11. A computer program product comprising computer instructions, characterized in that: When the computer instructions are executed by a processor, the steps of the method for determining the credit score of a loan user as described in any one of claims 1 to 7 are implemented.

Images