Market subject credit risk evaluation modeling method and device based on machine learning
Through machine learning-based methods, the risk index model and neural network risk evaluation model are trained, and the problem of insufficient intelligence and accuracy of credit risk evaluation of existing market entities is solved, and accurate assessment of credit risks of market entities and general credit are achieved. Establishment of risk assessment model.
Patent Information
- Application Number
- CN202411462167.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-18
- Publication Date
- 2025-02-11
AI Technical Summary
The existing credit risk assessment methods of market entities are not intelligent, have low accuracy, and lack of a general credit risk assessment index system and credit rating methods.
Using a machine learning-based method, automated credit risk assessment and credit rating are achieved by obtaining and preprocessing the credit risk assessment index data of market entities, and training the risk index model and neural network risk assessment model.
It realizes an accurate assessment of the credit risks of market entities, improves the accuracy of risk assessment, and provides a general credit risk assessment model suitable for all market entities.
Smart Images
Figure CN119417591A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of credit risk assessment, and in particular to a method and device for modeling credit risk assessment of market entities based on machine learning. Background Art
[0002] At present, with the development of economy, society and information technology, market vitality and social creativity have been effectively stimulated, the number of market players has grown rapidly, and new industries, new formats and new models have flourished, which has put forward higher requirements for market supervision resources, supervision capabilities and supervision intelligence. In order to effectively resolve the contradiction between heavy supervision tasks and insufficient supervision forces and further improve supervision efficiency, credit risk assessment of market players is an important part of building a new credit-based supervision mechanism, which plays an important role in improving the efficiency of supervision work, optimizing the allocation of supervision resources, and creating a market environment of honesty, trustworthiness and fair competition.
[0003] Expanding the concept and method of credit risk classification management of market entities to various business areas of market supervision and using the results of credit risk classification to scientifically allocate regulatory resources can improve the timeliness, accuracy and effectiveness of supervision, and promote fair competition and survival of the fittest through fair supervision. Credit risk assessment is an important issue in the field of market supervision, and plays a strong supporting role in optimizing the allocation of regulatory resources and improving regulatory effectiveness.
[0004] At present, the research on domestic corporate credit evaluation is deepening in various fields. The Delphi method, analytic hierarchy process, game analysis, entropy method, etc. are mainly used for comprehensive evaluation. The main problem is that it is subjective and relies more on expert experience. At the same time, the existing methods cannot automatically evaluate the credit risk of market entities, and the accuracy of risk evaluation is not high. There is no universal credit risk evaluation indicator system and credit rating method applicable to all market entities. Summary of the invention
[0005] The purpose of the embodiments of the present application is to provide a method and device for modeling credit risk assessment of market entities based on machine learning, so as to solve the problems that the credit risk assessment of market entities is not intelligent, the risk assessment accuracy is not high, and there is no universal credit risk assessment indicator system and credit rating method applicable to all market entities.
[0006] In a first aspect, the present application embodiment provides a market entity credit risk assessment modeling method based on machine learning, including:
[0007] Obtain relevant data on credit risk assessment indicators of market entities and pre-process the data;
[0008] Training a risk indicator model based on the preprocessed first data; wherein the first data includes a market subject credit risk evaluation indicator and market subject credit information data;
[0009] Based on the risk indicator model and the market subject credit risk evaluation index, the market subject credit information data is screened and initially classified to obtain second data; wherein the second data includes the indicator data and the market subject credit risk level data;
[0010] training a neural network risk assessment model based on the second data;
[0011] Based on the neural network risk assessment model, the credit risk level data of market entities is secondary classified to obtain the risk level prediction results.
[0012] In the above implementation process, relevant data of credit risk evaluation indicators of market entities are obtained and preprocessed; based on the preprocessed first data, a risk indicator model is trained; wherein, the first data includes credit risk evaluation indicators of market entities and credit information data of market entities; based on the risk indicator model and the credit risk evaluation indicators of market entities, the credit information data of market entities are screened and initially classified to obtain second data; wherein, the second data includes indicator data and credit risk grade data of market entities; based on the second data, a neural network risk evaluation model is trained; based on the neural network risk evaluation model, the credit risk grade data of market entities are secondary classified to obtain risk grade prediction results; credit risk evaluation indicators can be automatically extracted, and a universal credit risk evaluation model applicable to all market entities is trained to realize credit risk evaluation and credit rating, thereby improving the accuracy of risk evaluation.
[0013] Furthermore, the acquisition of relevant data of the credit risk assessment index of the market entity and preprocessing of the data include:
[0014] Obtain relevant data on credit risk assessment indicators of market entities;
[0015] Preprocessing is performed on relevant data of the credit risk assessment index of the market entity to obtain first data; wherein the preprocessing includes data cleaning, missing value and outlier processing, and data normalization processing.
[0016] In the above implementation process, the relevant data of the credit risk assessment indicators of market entities are preprocessed to ensure the quality and consistency of the data and prepare for subsequent model training.
[0017] Furthermore, the risk indicator model is trained based on the preprocessed first data, including:
[0018] Dividing the first data into a first training set and a first test set according to a set ratio;
[0019] Based on the first training set, the learning curve and grid search methods are used to traverse the hyperparameter combinations, and the optimal parameters are returned through 5-fold cross validation;
[0020] According to the first training set, a risk indicator model is trained;
[0021] The risk indicator model is verified based on the first test set, and the verification result is evaluated by the evaluation index; wherein the evaluation index includes AUC value, recall rate and precision rate.
[0022] In the above implementation process, the relevant data of the credit risk assessment indicators of market entities are divided into a first training set and a first test set, feature learning is performed on the first training set, and the model effect is verified on the first test set to obtain the final risk indicator model.
[0023] Furthermore, the step of training the risk indicator model according to the first training set includes:
[0024] Taking the credit risk evaluation index of market entities as the input feature of the model, combined with the optimal parameters, the credit risk level of market entities is predicted using the credit risk evaluation index of market entities;
[0025] The obtained credit risk level is encoded into a risk-free label, and the risk-free label is used as the output label of the model.
[0026] In the above implementation process, the risk indicator model is trained to output risk or no risk labels based on the relevant data of the credit risk evaluation indicators of market entities, so as to screen the credit risk evaluation indicators of market entities.
[0027] Furthermore, the second data is obtained by screening and initially classifying the credit information data of market entities based on the risk indicator model and the credit risk evaluation index of market entities, including:
[0028] Calculate the characteristic importance of credit risk assessment indicators of each market entity based on the training results of the risk indicator model;
[0029] Ranking the credit risk assessment indicators of various market entities based on their characteristic importance;
[0030] According to the ranking of the credit risk evaluation indicators of the corresponding market entities, the credit risk level data of the first market entity and the credit risk level data of the second market entity are obtained; wherein the credit risk level of the credit risk level data of the first market entity is less than or equal to the credit risk level of the credit risk level data of the second market entity;
[0031] The second data is obtained by integrating all the indicator data and the credit risk level data of the market entities;
[0032] The neural network risk assessment model is based on which the credit risk level data of market entities are secondary classified to obtain risk level prediction results, including:
[0033] Based on the neural network risk assessment model, the credit risk level data of the first market entity and the credit risk level data of the second market entity are secondary classified respectively;
[0034] Classify the credit risk level data of the first market entity as high risk level data, and further divide it into higher risk level and high risk level;
[0035] The second market entity credit risk level data is divided into four risk levels of market entity credit risk level data; among them, the four risk levels are low risk level, general risk level, higher risk level and high risk level.
[0036] In the above implementation process, the characteristic importance of each credit risk assessment indicator is calculated according to the training results of the risk indicator model, and the indicators are sorted according to the characteristic importance. The credit risk assessment indicators of market entities are screened according to the characteristic importance, and an initial classification and rating is performed. The results of the initial rating are subjected to a second classification and rating through a neural network risk assessment model to obtain credit risk level data of market entities at four risk levels, thereby improving the accuracy of risk assessment.
[0037] Further, the training of the neural network risk assessment model according to the second data includes:
[0038] Dividing the second data into a second training set and a second test set according to a set ratio;
[0039] Based on the second training set, the model training parameters are tuned using cross entropy loss to obtain the optimal parameter combination;
[0040] Training a neural network risk assessment model based on the second training set;
[0041] The neural network risk assessment model is predicted based on the second test set, and the prediction results are evaluated by evaluation indicators; wherein the evaluation indicators include AUC value, recall rate and precision rate.
[0042] In the above implementation process, a universal credit risk assessment model applicable to all market entities is trained to realize credit risk assessment and credit rating, thereby improving the accuracy of risk assessment.
[0043] Furthermore, the acquisition of relevant data of the credit risk assessment index of the market subject and the pre-processing of the data also includes:
[0044] Determine the credit risk evaluation indicators of market entities based on the risk influencing factors of market entities;
[0045] Among them, the market entity credit risk assessment indicators include: enterprise credit risk assessment indicators, individual business household credit risk assessment indicators, and farmers' professional cooperative credit risk assessment indicators.
[0046] In the above implementation process, the credit risk assessment indicators of different market entities can be automatically obtained, so as to facilitate subsequent training to obtain a universal credit risk assessment model applicable to all market entities.
[0047] In a second aspect, the present application embodiment provides a market entity credit risk assessment modeling device based on machine learning, comprising:
[0048] The data acquisition module is used to obtain relevant data of the credit risk evaluation indicators of market entities and pre-process the data;
[0049] A first training module is used to train a risk indicator model based on the preprocessed first data; wherein the first data includes a market subject credit risk evaluation indicator and market subject credit information data;
[0050] An indicator screening module, used to screen and initially classify the credit information data of market entities based on the risk indicator model and the credit risk evaluation indicators of market entities to obtain second data; wherein the second data includes indicator data and credit risk level data of market entities;
[0051] A second training module, used for training a neural network risk assessment model according to the second data;
[0052] The result acquisition module is used to perform secondary classification of the credit risk level data of market entities based on the neural network risk assessment model to obtain risk level prediction results.
[0053] In a third aspect, an embodiment of the present application provides an electronic device, including:
[0054] A processor, a memory and a bus, wherein the processor is connected to the memory via the bus, and the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, they are used to implement the market entity credit risk assessment modeling method based on machine learning as described above.
[0055] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a server, implements the market entity credit risk assessment modeling method based on machine learning as described above. BRIEF DESCRIPTION OF THE DRAWINGS
[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.
[0057] Figure 1 A flowchart of a market entity credit risk assessment modeling method based on machine learning provided in an embodiment of the present application;
[0058] Figure 2 It is a structural schematic diagram of a market subject credit risk assessment modeling device based on machine learning provided in an embodiment of the present application;
[0059] Figure 3 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0060] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.
[0061] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.
[0062] In recent years, with the development of economy, society and information technology, market vitality and social creativity have been effectively stimulated, the number of market players has grown rapidly, and new industries, new formats and new models have flourished, which has put forward higher requirements for market supervision resources, supervision capabilities and supervision intelligence. In order to effectively resolve the contradiction between heavy supervision tasks and insufficient supervision forces and further improve supervision efficiency, credit risk assessment of market players is an important part of building a new supervision mechanism based on credit, which plays an important role in improving the efficiency of supervision work, optimizing the allocation of supervision resources, and creating a market environment of honesty, trustworthiness and fair competition.
[0063] Scientifically establish a market entity credit risk classification index system, dynamic management indicators and weights, comprehensively use various types of enterprise-related information, rely on information technology to implement timely and automatic classification, and objectively reflect the credit risk status of market entities.
[0064] The existing credit risk classification mainly adopts a comprehensive evaluation method, determines a risk assessment indicator system for professional industry fields, and calculates risk scores by weighting and scoring indicators to conduct risk assessment and classification.
[0065] Indicator processing: Building a scientific indicator system is the key to the classification management of credit risk of market entities. In combination with the actual work in the field of market supervision, through various technical means such as indicator screening, indicator construction, and indicator preprocessing, the data is integrated, processed, classified, and calculated, and the indicator monitoring scope is increased through statistical analysis, making the evaluation indicators more comprehensive and objective. At the same time, the new indicator system is open to meet the needs of continuous optimization based on data resources and application conditions.
[0066] (1) Indicator screening principles
[0067] The credit risk classification management and evaluation system for market entities should be scientific, practical, and operational. Therefore, the correct guiding principles and screening indicator system should be determined to achieve the desired purpose.
[0068] 1) Hierarchical principle. The selection of indicators should cover the elements of risk in the field of enterprise market supervision from different levels and directions as much as possible, so as to comprehensively reflect the credit, compliance with laws and regulations and business conditions of the enterprise.
[0069] 2) The principle of intensiveness. The selected indicators should have sufficient information integration capabilities. Too many indicators will be cumbersome, and too few indicators will not be able to comprehensively evaluate the various risk conditions of enterprises in the field of market supervision.
[0070] 3) Principle of comparability. The credit risk classification indicators of market supervision enterprises should have universal significance. The setting of evaluation indicators should take into account both vertical comparability, dynamically reflecting the development process and internal change rules of corporate credit risk, and horizontal comparison, so as to achieve comparability of risks of different enterprises.
[0071] 4) Scientific principle. As a systematic project, the classification of credit risk of market supervision enterprises must be formed by organic combination of various indicators, which should neither be repeated nor contradictory. This is the key to whether the supervision risk assessment indicator system is scientific.
[0072] 5) Principle of practicality. On the one hand, the selected indicator data should be relatively easy to obtain. On the other hand, the indicator selection should not be too complicated. The principle of comprehensiveness and the principle of practicality should be unified with key and representative indicators. At the same time, the selection of indicators should not only conform to the actual business situation in the field of market supervision, but also refer to some common international and domestic practices, taking into account both scientificity and practicality.
[0073] Indicator construction technology: converting raw data items into indicators. Raw indicators refer to indicators that are directly established using the collected indicator items. Based on modeling needs, raw indicators also need to be appropriately transformed. Specific transformation methods include at least the following:
[0074] (1) Continuous data items are converted into continuous indicators required for modeling.
[0075] Continuous variables are still converted into continuous variables, usually to make the data comparable, or to correct the distribution of the data, or to make it easier to use, such as standardizing the tax payments and assets of enterprises.
[0076] (2) Continuous data items are converted into discrete indicators.
[0077] Continuous data items are converted into discrete indicators, usually to facilitate better understanding and application from a business perspective, and sometimes to accommodate certain classification algorithms that cannot directly handle continuous variables.
[0078] The discretization of continuous attributes is to set a number of discrete division points within the value range of a specific continuous attribute, divide the value range of the attribute into a number of discrete intervals, and finally use different symbols or integer values to represent the attribute values falling in each sub-interval. The discretization process of continuous attributes is essentially the process of dividing the space formed by conditional attributes using selected breakpoints.
[0079] The purpose of discretization of continuous values is to simplify the data structure. Data usually has several types: nominal, discrete and continuous. Data in real life often comes from composite types. For example, many data are composites of numerical and discrete types. This requires a unified data type for data processing. By dividing the attribute value domain into intervals, data discretization technology can be used to reduce the number of given continuous attribute values. Discrete attributes are closer to knowledge-level expressions than continuous attributes. For users and experts, discrete attributes are easier to understand, use and interpret. Discretization methods are often used as tools for data mining, and the results obtained using discrete attributes (decision trees, induction rules) are often more compact, concise and accurate, so their results are also easier to check, compare, use and reuse.
[0080] In addition, some classification learning algorithms can only handle discrete values. Discretization is the process of quantizing continuous attributes, which can significantly expand the application range of many learning algorithms. The common normal assumption is for continuous variables, and discretization reduces the dependence on distribution assumptions, so discrete data is sometimes more effective.
[0081] (3) Discrete indicators are converted into continuous indicators.
[0082] The most common situation of converting discrete indicators into continuous indicators is that for binary marker variables, the values are 0 and 1 respectively, so that the original data becomes a continuous indicator between 0 and 1.
[0083] (4) Discrete indicator items are converted into discrete indicators required for modeling.
[0084] Discrete indicator items are still converted into discrete indicators. In most cases, there are too many categories of discrete indicators, and they need to be appropriately merged during modeling.
[0085] Expanding the concept and method of credit risk classification management of market entities to various business areas of market supervision and using the results of enterprise credit risk classification to scientifically allocate regulatory resources can improve the timeliness, accuracy and effectiveness of supervision, and promote fair competition and survival of the fittest through fair supervision. Credit risk assessment is an important issue in the field of market supervision, and plays a strong supporting role in optimizing the allocation of regulatory resources and improving regulatory effectiveness.
[0086] At present, the research on domestic corporate credit evaluation is deepening in various fields. The Delphi method, analytic hierarchy process, game analysis, entropy method, etc. are mainly used for comprehensive evaluation. The main problem is that it is subjective and relies more on expert experience. At the same time, the existing methods cannot automatically evaluate the credit risk of market entities, and the accuracy of risk evaluation is not high. There is no credit risk evaluation indicator system and credit rating method applicable to all market entities.
[0087] Please see Figure 1 , Figure 1 A flow chart of a method for modeling credit risk assessment of a market subject based on machine learning provided in an embodiment of the present application. The method for modeling credit risk assessment of a market subject based on machine learning includes:
[0088] 100. Obtain relevant data on credit risk assessment indicators of market entities and pre-process the data.
[0089] Optionally, before obtaining the relevant data of the credit risk evaluation index of the market entity and preprocessing the data, the method further includes:
[0090] Based on the risk influencing factors of market entities, the credit risk evaluation indicators of market entities are determined; wherein the credit risk evaluation indicators of market entities include: corporate credit risk evaluation indicators, individual business credit risk evaluation indicators, and farmers' professional cooperative credit risk evaluation indicators.
[0091] Thus, the credit risk assessment indicators of different market players can be automatically obtained, which facilitates subsequent training to obtain a universal credit risk assessment model applicable to all market players.
[0092] 110. Obtain relevant data on credit risk assessment indicators of market entities, where the data includes credit risk assessment indicators of market entities and credit information data of market entities.
[0093] Specifically, the market subject credit data used in the method of the embodiment of the present application is the market subject credit data generated within the market supervision system, as well as various information generated by multiple departments performing work. In addition, the market subject credit data adopts unified standards and specifications in terms of data collection caliber and collection method, has a complete information resource integration mechanism, and has a unified information resource integration and exchange mechanism to meet the needs of the method of the embodiment of the present application.
[0094] Specifically, relevant data on credit risk assessment indicators of market entities can be obtained through a variety of channels, such as public channels, third-party data providers, internal databases of market supervision, and data obtained through data exchange with other departments. The data acquisition method is accurate and fast, which improves the overall modeling accuracy and efficiency. Among them, the credit risk assessment indicators of market entities can be selected based on industry characteristics, risk assessment objectives and regulatory requirements, which may include but are not limited to financial indicators (such as asset-liability ratio, profit margin), operating indicators (such as tax and social security, trademarks and patents), credit history (such as illegal records, judicial proceedings), industry-specific indicators (such as industry risk assessment, policy impact), etc. Among them, the credit risk level data of market entities is usually the result of risk classification of market entities based on a series of assessment standards and indicators, for example, in the form of letters (such as AD level, A level has the lowest risk) or text descriptions (such as low risk, general risk, higher risk, high risk).
[0095] 120. Preprocess the relevant data of the credit risk assessment indicators of market entities to obtain first data; wherein the preprocessing includes data cleaning, missing value and outlier processing, and data normalization.
[0096] Understandably, the data is cleaned, duplicate data is removed, missing values are processed, outliers are detected and processed to ensure the accuracy and completeness of the data; the data is normalized and data from different sources are converted into a unified format for subsequent analysis.
[0097] Optionally, it may also include data integration: integrating market entity credit risk assessment indicators and market entity credit information data to form a complete data set, which can be achieved, for example, through data matching and data merging.
[0098] 200. Train a risk indicator model based on the preprocessed first data; wherein the first data includes a market entity credit risk assessment indicator and market entity credit information data.
[0099] Optionally, the risk indicator model adopts an XGBoost model.
[0100] 210. Divide the first data into a first training set and a first test set according to a set ratio.
[0101] Exemplarily, the ratio is set to 4:1, that is, the first data (market entity credit risk assessment indicators and market entity credit information data) is divided into a first training set and a first test set in a ratio of 4:1; thus, feature learning is performed on the first training set, and the model effect is verified on the first test set.
[0102] There are many methods for dividing the first training set and the first test set, such as random division, hierarchical division, and time sequence division.
[0103] Optionally, before dividing the training set and the test set, shuffle the data to ensure the randomness and fairness of the division results; after the division is completed, the training set and the test set need to be saved as independent files or database tables for use in subsequent training and testing processes; after the division is completed, the training set and the test set can also be verified to ensure that their key features such as data distribution and category ratio meet expectations.
[0104] 220. Based on the first training set, the learning curve and grid search method are used to traverse the hyperparameter combination, and the optimal parameters are returned through 5-fold cross validation.
[0105] Specifically, the max_depth, n_estimators, gamma and other parameters in the XGBoost model are tuned, and the hyperparameter combinations are traversed using the learning curve and grid search methods. The optimal parameters are returned through 5-fold cross validation, thereby optimizing the performance of the XGBoost model.
[0106] 230. Train a risk indicator model based on the first training set.
[0107] 240. Verify the risk indicator model based on the first test set, and evaluate the verification result through the evaluation indicator.
[0108] Among them, the evaluation indicators include AUC value (Area Under the Curve, which is the area under the ROC curve, used to evaluate the performance of the classifier under different thresholds), recall rate (the ratio of correctly identified positive samples to the total number of positive samples) and precision rate (the ratio of correctly identified positive samples to the number of predicted positive samples).
[0109] Specifically, the performance of the model on the first test set is analyzed based on the calculated AUC value, recall rate, and precision rate. The closer the AUC value is to 1, the better the performance of the model; the higher the recall rate, the more positive samples the model recognizes; the higher the precision rate, the higher the proportion of samples predicted by the model as positive that are actually positive.
[0110] As described above, the embodiment of the present application divides the relevant data of the credit risk assessment index of the market entity into a first training set and a first test set, performs feature learning on the first training set, verifies the model effect on the first test set, and obtains the final risk indicator model.
[0111] Based on the above implementation, the embodiment of the present application can also be embodied as follows: the step of training the risk indicator model according to the first training set includes:
[0112] 231. Taking the credit risk evaluation index of market entities as the input feature of the model, combined with the optimal parameters, the credit risk level of market entities is predicted using the credit risk evaluation index of market entities;
[0113] Specifically, the credit risk evaluation indicators of market entities in the first training set are used as the input features of the model, combined with the obtained optimal parameters of the model to train the risk indicator model, and the credit risk level of the market entity output by the model is predicted using the credit risk evaluation indicators of market entities.
[0114] 232. Encode the obtained credit risk level into a risk-free label, and use the risk-free label as the output label of the model.
[0115] Specifically, when training the XGBoost model, the credit risk level of the market subject is encoded into a label of "risk or not". The label of "risk or not" is used as the output label y of XGBoost, and the credit risk evaluation index data x of the market subject is used as the input feature of XGBoost, so that the market subject can use the credit risk evaluation index to predict whether the enterprise has risk or not.
[0116] Among them, the obtained credit risk level of the market entity is encoded into a label of "risk or not", and can be sorted from high to low according to the credit risk level of the market entity, and the "risk or not" division ratio is set according to the sorting result. For example, the division ratio can be 5:5, 7:3, etc.
[0117] By setting the credit risk evaluation index of market entities as input features, setting the "risk or not" label obtained by the credit risk level of market entities as the output label, and combining the optimal parameters of the model, the precision and accuracy of the trained risk indicator model can be higher.
[0118] As described above, the embodiment of the present application trains a risk indicator model to output a risk-free label based on relevant data of the credit risk evaluation indicators of market entities, so as to screen the credit risk evaluation indicators of market entities.
[0119] 300. Based on the risk indicator model and the market entity credit risk evaluation index, screen and initially classify the market entity credit information data to obtain second data; wherein the second data includes indicator data and market entity credit risk level data.
[0120] 310. Calculate the characteristic importance of each credit risk assessment indicator based on the training results of the risk indicator model.
[0121] It can be understood that feature importance refers to the degree of influence of each input feature on the model output (i.e. credit risk level) during the model prediction process.
[0122] Specifically, the characteristic importance of each credit risk assessment indicator is calculated. It can be understood that there are many ways to calculate the characteristic importance, and the embodiments of the present application do not specifically limit this.
[0123] 320. Sort each credit risk assessment indicator based on its characteristic importance.
[0124] Specifically, the various credit risk assessment indicators are ranked according to the standardized feature importance values. Optionally, the ranking results may be presented in a list format, listing the names of the various features and the corresponding feature importance values, or using visualization tools such as bar charts and column charts to present the feature importance and ranking.
[0125] 330. According to the ranking of the credit risk evaluation indicators of corresponding market entities, the credit risk level data of the first market entity and the credit risk level data of the second market entity are obtained; wherein, the credit risk level of the credit risk level data of the first market entity is less than or equal to the credit risk level of the credit risk level data of the second market entity.
[0126] Specifically, an initial classification and evaluation is performed on the credit information data of market entities to obtain credit information data of two risk levels, namely, first market entity credit risk level data and second market entity credit risk level data; illustratively, the first market entity credit risk level data is classified as high risk level data, which is further divided into relatively high risk level and high risk level.
[0127] 340. The second data is obtained by integrating all indicator data and credit risk level data of market entities.
[0128] Exemplarily, the credit risk assessment indicators are screened according to the training results of the XGBoost model. Specifically, the feature importance of each credit risk assessment indicator is calculated according to the training results of the XGBoost model, and each credit risk assessment indicator is sorted according to the feature importance, and data with a higher risk level and data with a high risk level are screened according to the feature importance.
[0129] 400. Train a neural network risk assessment model based on the second data.
[0130] Optionally, the indicator data and the credit risk level data of the market entities screened by the XGBoost model are input into the BP neural network to train the BP neural network risk assessment model.
[0131] 410. Divide the second data into a second training set and a second test set according to a set ratio.
[0132] Exemplarily, the ratio is set to 7:3, that is, the second data (indicator data and market entity credit risk level data) is divided into a second training set and a second test set in a ratio of 7:3; thus, feature learning is performed on the second training set, and the model effect is verified on the second test set.
[0133] 420. Based on the second training set, the model training parameters are tuned using cross entropy loss to obtain the optimal parameter combination.
[0134] Specifically, based on the gradient obtained by calculating the cross entropy loss, an optimization algorithm (such as gradient descent, Adam, etc.) is used to update the model parameters to obtain the optimal parameter combination to reduce the loss and enable the model to more accurately divide future ratings.
[0135] 430. Train a neural network risk assessment model based on the second training set.
[0136] Specifically, the risk indicator model is trained based on the second training set and the obtained optimal model parameter combination.
[0137] 440. Predict the neural network risk assessment model based on the second test set, and evaluate the prediction results through evaluation indicators; wherein the evaluation indicators include AUC value, recall rate and precision rate.
[0138] Optionally, the model may be optimized based on the evaluation results. Exemplarily, based on specific requirements, the model may be optimized using Bayesian optimization or genetic algorithm.
[0139] For example, if the recall rate is low, you can try to adjust the threshold of the model or increase the complexity of the model; if the precision rate is low, you can try to introduce more features or perform feature selection.
[0140] In the above implementation process, a universal credit risk assessment model applicable to all market entities is trained to realize credit risk assessment and credit rating, thereby improving the accuracy of risk assessment.
[0141] 500. Based on the neural network risk assessment model, the credit risk level data of market entities is secondary classified to obtain the risk level prediction results.
[0142] 510. Based on the neural network risk assessment model, the credit risk level data of the first market entity and the credit risk level data of the second market entity are secondary classified respectively.
[0143] 520. The credit risk level data of the first market entity shall be classified as high-risk level data, and further divided into higher risk level and high risk level.
[0144] 530. The credit risk level data of the second market entity is divided into four risk levels of market entity credit risk level data; among which, the four risk levels are low risk level, general risk level, relatively high risk level and high risk level.
[0145] Exemplarily, the credit risk level data of the first market entity is classified secondary into C and D (higher risk level and high risk level) data; in addition, the credit risk level data of the second market entity is divided into four risk levels, namely A, B, C, and D (low risk level, general risk level, higher risk level, and high risk level) data.
[0146] As described above, the embodiment of the present application calculates the characteristic importance of each credit risk assessment indicator based on the training results of the risk indicator model, sorts them by characteristic importance, screens the credit risk assessment indicators of market entities based on characteristic importance, and performs initial classification and rating. The results of the initial rating are subjected to a second classification and rating through a neural network risk assessment model to obtain credit risk level data of market entities at four risk levels, thereby improving the accuracy of risk assessment.
[0147] It can be understood that the comprehensive evaluation method of the credit risk of market entities in the embodiment of the present application consists of two parts: a credit risk indicator model and a risk evaluation model. By collecting data related to market entities such as enterprise credit information disclosure system data and cross-departmental data, and based on the existing foundation and actual needs, the enterprise credit risk evaluation indicator model, the individual business household credit risk evaluation indicator model, and the farmer professional cooperative credit risk evaluation indicator model are constructed for the three types of market entities, namely, enterprises, individual businesses, and farmers professional cooperatives. Thus, by inputting the credit risk indicator data of the market entity (enterprise, individual business household or farmer professional cooperative) into the corresponding risk evaluation indicator model, the risk level prediction result of the corresponding market entity can be obtained.
[0148] Compared with the existing enterprise credit risk classification using a balanced scorecard for credit scoring, the market entity credit risk comprehensive evaluation method of the embodiment of the present application uses big data technology to automatically learn the risk characteristics of market entities, reduces the dependence of indicator weights and scoring in the balanced scorecard on expert experience, improves the accuracy of risk evaluation classification, and at the same time achieves coverage of all entities including enterprises, individual industrial and commercial households, and farmers' professional cooperatives.
[0149] The embodiment of the present application is applicable to the risk assessment of all market entities. At the same time, the embodiment of the present application can automatically extract credit risk assessment indicators, thereby improving the accuracy of risk assessment.
[0150] The method of the embodiment of the present application covers all market entities, and divides all market entities into four categories according to their risk levels: high risk, relatively high risk, general risk, and low risk. The concept and method of credit risk classification management of market entities can be extended to various business areas of market supervision, and the risk classification results can be applied to various business areas such as double random supervision, special supervision, and key field supervision. The credit risk classification results of market entities are used to scientifically allocate supervision resources, increase the inspection frequency and inspection intensity for high-risk entities, and reduce the inspection frequency for low-risk entities, so as to improve the timeliness, accuracy, and effectiveness of supervision, so that supervision is "ubiquitous" for lawbreakers and dishonest people, and "does not disturb" honest and law-abiding people, so as to promote fair competition and survival of the fittest with fair supervision. At the same time, the credit risk evaluation results can be shared with the "Internet + Supervision" system to promote the construction of the credit system; the credit risk evaluation results of various market entities are used to build credit files, encourage enterprises to operate in compliance with laws and regulations, and improve credit levels.
[0151] The above steps are not to be performed in a strict order as described in the numbers, but should be understood as an overall solution.
[0152] In the second aspect, based on the above embodiments, Figure 2 A schematic diagram of a machine learning-based credit risk assessment modeling device for market entities provided in an embodiment of the present application. Figure 2 The market entity credit risk assessment modeling device based on machine learning provided in this embodiment specifically includes: a data acquisition module 201, a first training module 202, an indicator screening module 203, a second training module 204 and a result acquisition module 205.
[0153] Among them, the data acquisition module 201 is used to obtain relevant data of the credit risk evaluation index of the market entity and pre-process the data; the first training module 202 is used to train the risk indicator model based on the pre-processed first data; wherein the first data includes the credit risk evaluation index of the market entity and the credit information data of the market entity; the indicator screening module 203 is used to screen and initially classify the credit information data of the market entity based on the risk indicator model and the credit risk evaluation index of the market entity to obtain the second data; wherein the second data includes the indicator data and the credit risk level data of the market entity; the second training module 204 is used to train the neural network risk evaluation model according to the second data; the result acquisition module 205 is used to perform secondary classification on the credit risk level data of the market entity based on the neural network risk evaluation model to obtain the risk level prediction result.
[0154] As described above, the embodiment of the present application obtains relevant data of the credit risk evaluation index of the market entity and pre-processes the data; trains to obtain a risk indicator model based on the pre-processed first data; based on the risk indicator model and the credit risk evaluation index of the market entity, screens and initially classifies the credit information data of the market entity to obtain second data; trains a neural network risk evaluation model based on the second data; based on the neural network risk evaluation model, performs secondary classification on the credit risk level data of the market entity to obtain a risk level prediction result; can automatically extract credit risk evaluation indicators, train to obtain a universal credit risk evaluation model applicable to all market entities, realize credit risk evaluation and credit rating, and improve the accuracy of risk evaluation.
[0155] The market entity credit risk assessment modeling device based on machine learning provided in the embodiment of the present application can be used to execute the market entity credit risk assessment modeling method based on machine learning provided in the above-mentioned embodiment, and has corresponding functions and beneficial effects.
[0156] On the third aspect, the embodiments of the present application also provide an electronic device that can integrate the market entity credit risk assessment modeling device based on machine learning provided in the embodiments of the present application. Figure 3 Schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Figure 3 The electronic device includes: an input device 33, an output device 34, a memory 32 and one or more processors 31; the memory 32 is used to store one or more programs; when the one or more programs are executed by the one or more processors 31, the one or more processors 31 implement the market subject credit risk assessment modeling method based on machine learning as provided in the above embodiment. The input device 33, the output device 34, the memory 32 and the processor 31 can be connected by a bus or other means, Figure 3 The example of connecting through bus is taken in the following.
[0157] The processor 31 executes various functional applications and data processing of the device by running the software programs, instructions and modules stored in the memory 32, that is, realizes the above-mentioned market entity credit risk assessment modeling method based on machine learning.
[0158] The electronic device provided above can be used to execute the market entity credit risk assessment modeling method based on machine learning provided in the above embodiment, and has corresponding functions and beneficial effects.
[0159] In a fourth aspect, an embodiment of the present application also provides a computer-readable storage medium, which includes a stored computer program; wherein, when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the market entity credit risk assessment modeling method based on machine learning as described above, and can achieve the same beneficial effects.
[0160] Of course, the storage medium containing computer executable instructions provided in an embodiment of the present application is not limited to the market entity credit risk assessment modeling method based on machine learning as described above, and can also execute related operations in the market entity credit risk assessment modeling method based on machine learning provided in any embodiment of the present application.
[0161] In a fifth aspect, the embodiments of the present application also provide a computer program product, and the methods described in the various embodiments of the present application can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instruction is loaded and executed on a computer, the process or function described in the various embodiments of the present application is executed in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user device, a core network device, an OAM (Open Application Model) or other programmable device.
[0162] The computer program or instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer program or instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired or wireless means. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape; it may also be an optical medium, such as a digital video disk; it may also be a semiconductor medium, such as a solid-state drive. The computer-readable storage medium may be a volatile or non-volatile storage medium, or may include both volatile and non-volatile types of storage media.
[0163] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.
[0164] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.
[0165] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for enabling an electronic device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.
[0166] The above description is only an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings.
[0167] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
[0168] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.
Claims
1. A market subject credit risk assessment modeling method based on machine learning, characterized in that: The method comprises: Obtain relevant data on credit risk assessment indicators of market entities and pre-process the data; A risk indicator model is obtained by training according to the preprocessed first data; wherein the first data includes a market subject credit risk evaluation indicator and market subject credit information data; Based on the risk indicator model and the market subject credit risk evaluation index, the market subject credit information data is screened and initially classified to obtain second data; wherein the second data includes the indicator data and the market subject credit risk level data; training a neural network risk assessment model based on the second data; Based on the neural network risk assessment model, the credit risk level data of market entities is secondary classified to obtain the risk level prediction results.
2. The method for market subject credit risk assessment modeling based on machine learning according to claim 1 is characterized in that: The obtaining of relevant data on the credit risk assessment indicators of market entities and preprocessing of the data include: Obtain relevant data on credit risk assessment indicators of market entities; Preprocessing is performed on relevant data of the credit risk assessment index of the market entity to obtain first data; wherein the preprocessing includes data cleaning, missing value and outlier processing, and data normalization processing.
3. The market subject credit risk assessment modeling method based on machine learning according to claim 1 is characterized in that: The step of training a risk indicator model based on the preprocessed first data includes: Dividing the first data into a first training set and a first test set according to a set ratio; Based on the first training set, the learning curve and grid search methods are used to traverse the hyperparameter combinations, and the optimal parameters are returned through 5-fold cross validation; According to the first training set, a risk indicator model is trained; The risk indicator model is verified based on the first test set, and the verification result is evaluated by the evaluation index; wherein the evaluation index includes AUC value, recall rate and precision rate.
4. The method for market subject credit risk assessment modeling based on machine learning according to claim 3 is characterized in that: The step of training the risk indicator model according to the first training set includes: Taking the credit risk evaluation index of market entities as the input feature of the model, combined with the optimal parameters, the credit risk level of market entities is predicted using the credit risk evaluation index of market entities; The obtained credit risk level is encoded into a risk-free label, and the risk-free label is used as the output label of the model.
5. The market subject credit risk assessment modeling method based on machine learning according to claim 1 is characterized in that: The method of screening and initially classifying the credit information data of market entities based on the risk indicator model and the credit risk evaluation index of market entities to obtain the second data includes: Calculate the characteristic importance of credit risk assessment indicators of each market entity based on the training results of the risk indicator model; Ranking the credit risk assessment indicators of various market entities based on their characteristic importance; According to the ranking of the credit risk evaluation indicators of the corresponding market entities, the credit risk level data of the first market entity and the credit risk level data of the second market entity are obtained; wherein the credit risk level of the credit risk level data of the first market entity is less than or equal to the credit risk level of the credit risk level data of the second market entity; The second data is obtained by integrating all the indicator data and the credit risk level data of the market entities; The neural network risk assessment model is based on which the credit risk level data of market entities are secondary classified to obtain risk level prediction results, including: Based on the neural network risk assessment model, the credit risk level data of the first market entity and the credit risk level data of the second market entity are secondary classified respectively; Classify the credit risk level data of the first market entity as high risk level data, and further divide it into higher risk level and high risk level; The second market entity credit risk level data is divided into four risk levels of market entity credit risk level data; among them, the four risk levels are low risk level, general risk level, higher risk level and high risk level.
6. The market subject credit risk assessment modeling method based on machine learning according to claim 1 is characterized in that: The step of training the neural network risk assessment model according to the second data includes: Dividing the second data into a second training set and a second test set according to a set ratio; Based on the second training set, the model training parameters are tuned using cross entropy loss to obtain the optimal parameter combination; Training a neural network risk assessment model based on the second training set; The neural network risk assessment model is predicted based on the second test set, and the prediction results are evaluated by evaluation indicators; wherein the evaluation indicators include AUC value, recall rate and precision rate.
7. The market subject credit risk assessment modeling method based on machine learning according to claim 1 is characterized in that: The step of obtaining the relevant data of the credit risk evaluation index of the market subject and preprocessing the data also includes: Determine the credit risk evaluation indicators of market entities based on the risk influencing factors of market entities; Among them, the market entity credit risk assessment indicators include: enterprise credit risk assessment indicators, individual business household credit risk assessment indicators, and farmers' professional cooperative credit risk assessment indicators.
8. A market subject credit risk assessment modeling device based on machine learning, characterized in that: include: The data acquisition module is used to obtain relevant data of the credit risk evaluation indicators of market entities and pre-process the data; A first training module is used to train a risk indicator model based on the pre-processed first data; wherein the first data includes a market subject credit risk evaluation indicator and market subject credit information data; An indicator screening module, used to screen and initially classify the credit information data of market entities based on the risk indicator model and the credit risk evaluation indicators of market entities to obtain second data; wherein the second data includes indicator data and credit risk level data of market entities; A second training module, used for training a neural network risk assessment model according to the second data; The result acquisition module is used to perform secondary classification of the credit risk level data of market entities based on the neural network risk assessment model to obtain risk level prediction results.
9. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the processor is connected to the memory via the bus, and the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, they are used to implement the market entity credit risk assessment modeling method based on machine learning as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a server, implements the market entity credit risk assessment modeling method based on machine learning as described in any one of claims 1 to 7.