Enterprise credit granting processing method and device

Abstract

The invention discloses an enterprise credit granting processing method and device, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining the comprehensive data of a to-be-tested enterprise; the enterprise to be tested belongs to a large-scale or medium-scale enterprise; inputting the comprehensive data into a pre-constructed admission model, and outputting a result whether admission is performed or not; when the result is admission, inputting the comprehensive data of the to-be-tested enterprise into a pre-constructed credit analysis model, and outputting a credit risk prediction value; the credit granting analysis model is used for measuring and calculating enterprise financial risk, an enterprise contract fulfillment capability value and an enterprise operation turnover capability value; and adjusting the credit granting risk prediction value by using the medical and invasive financial associated data of the to-be-tested enterprise in the target bank, and outputting a credit granting evaluation result of the target bank for the to-be-tested enterprise. According to the invention, the credit granting processing efficiency of large and medium-sized science and technology enterprises can be improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an enterprise credit granting processing method and device. Background Art

[0002] This section aims to provide background or context for the embodiments of the present invention described in the claims. The descriptions herein are not admitted to be prior art merely because they are included in this section.

[0003] Enterprises can be classified into large, medium, small, and micro-sized according to their scale. Among them, for small and micro-sized enterprises, there are many credit granting processing models in the banking industry, and most banks can quickly respond to customers' credit granting processing needs through online risk prediction. For large and medium-sized enterprise customers, there is no unified credit granting processing model, and it is necessary to rely heavily on offline manual expert experience judgment, resulting in a large amount of credit approval work and a low digital level.

[0004] The main disadvantages of the existing technology for credit risk prediction of large and medium-sized enterprises are as follows:

[0005] First, the model scalability is weak. The current method has a long calculation process and poor scalability. Large and medium-sized technology-based enterprises are widely distributed in industries and have rapid technological iterations. The existing technology has limited flexibility and many application limitations, and cannot meet the high-intensity model expansion requirements.

[0006] Second, the maintenance cost is relatively high. When the existing technology is used for credit granting processing, it involves multiple business sub-modules, resulting in great pressure on operation and maintenance costs, labor costs, monitoring costs, etc. Technology-based enterprises have rapid business changes and diverse financing methods, and have higher requirements for loan products. The high maintenance cost of the current technology cannot match the actual financial service needs of large and medium-sized technology-based enterprises. Summary of the Invention

[0007] Embodiments of the present invention provide an enterprise credit granting processing method for using artificial intelligence technology to conduct credit evaluation processing on large and medium-sized technology-based enterprises, reduce labor costs and maintenance costs, improve model scalability, and improve credit granting processing efficiency. The method includes:

[0008] Obtain comprehensive data of the enterprise to be measured; the enterprise to be measured belongs to a large-scale enterprise or a medium-scale enterprise divided according to preset indicators; the comprehensive data includes financial data, industrial chain data, and scientific and technological innovation ability data;

[0009] Input the comprehensive data of the enterprise to be measured into a pre-constructed access model, and output the result of whether to access; the access model is pre-trained on a machine learning model by using the comprehensive data of historical enterprises that have been accessed.

[0010] When the result is for admission, input the comprehensive data of the enterprise to be tested into a pre-constructed credit analysis model, and output the predicted credit risk value of the enterprise to be tested; the credit analysis model includes a first analysis module, a second analysis module and a third analysis module. The first analysis module is used to predict the financial risks of the enterprise based on the extreme gradient boosting regression model. The second analysis module is used to calculate the enterprise's contract performance ability value. The third analysis module is used to calculate the enterprise's operation turnover ability value;

[0011] Use the science and technology innovation financial association data of the enterprise to be tested in the target bank to adjust the predicted credit risk value of the enterprise to be tested, and output the credit evaluation result of the target bank for the enterprise to be tested; the science and technology innovation financial association data includes the information data of the existing cooperative banks of the enterprise to be tested, the science and technology innovation evaluation data and credit data of the enterprise to be tested in the target bank.

[0012] The embodiment of the present invention also provides an enterprise credit processing device, which is used to use artificial intelligence technology to conduct credit evaluation processing on large and medium-sized science and technology enterprises, reduce labor costs and maintenance costs, improve model scalability, and improve credit processing efficiency. The device includes:

[0013] A data acquisition module, which is used to acquire the comprehensive data of the enterprise to be tested; the enterprise to be tested belongs to a large-scale enterprise or a medium-scale enterprise divided according to preset indicators; the comprehensive data includes financial data, industrial chain data, and science and technology innovation ability data;

[0014] An admission judgment module, which is used to input the comprehensive data of the enterprise to be tested into a pre-constructed admission model and output the result of whether to admit; the admission model is pre-trained on a machine learning model using the comprehensive data of historical admitted enterprises;

[0015] A credit analysis module, which is used to, when the result is for admission, input the comprehensive data of the enterprise to be tested into a pre-constructed credit analysis model, and output the predicted credit risk value of the enterprise to be tested; the credit analysis model includes a first analysis module, a second analysis module and a third analysis module. The first analysis module is used to predict the financial risks of the enterprise based on the extreme gradient boosting regression model. The second analysis module is used to calculate the enterprise's contract performance ability value. The third analysis module is used to calculate the enterprise's operation turnover ability value;

[0016] An adjustment module, which is used to use the science and technology innovation financial association data of the enterprise to be tested in the target bank to adjust the predicted credit risk value of the enterprise to be tested, and output the credit evaluation result of the target bank for the enterprise to be tested; the science and technology innovation financial association data includes the information data of the existing cooperative banks of the enterprise to be tested, the science and technology innovation evaluation data and credit data of the enterprise to be tested in the target bank.

[0017] An embodiment of the present invention further provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the above-mentioned enterprise credit-granting processing method is implemented.

[0018] An embodiment of the present invention further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned enterprise credit-granting processing method is implemented.

[0019] An embodiment of the present invention further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the above-mentioned enterprise credit-granting processing method is implemented.

[0020] In an embodiment of the present invention, for large-scale enterprises or medium-scale enterprises, comprehensive data is obtained. First, a machine learning model is used to determine whether the enterprise to be tested is eligible. When the result is eligible, a pre-constructed credit analysis model is used to process and obtain the credit risk prediction value of the enterprise to be tested. The credit analysis model includes a first analysis module for predicting the financial risks of the enterprise, a second analysis module for calculating the enterprise's contract performance ability value, and a third analysis module for calculating the enterprise's operation turnover ability value. Finally, using the science and technology innovation financial association data of the enterprise to be tested in the target bank, the credit risk prediction value of the enterprise to be tested is adjusted, and the credit assessment result of the target bank for the enterprise to be tested is output. Compared with the prior art, in the embodiment of the present invention, the credit analysis and processing of large and medium-sized technology-based enterprises are carried out through artificial intelligence technology. During the processing, multiple small business modules are not involved, and the prediction and judgment are directly carried out through two intelligent analysis models, which releases manpower and reduces the dimensionality cost. The model can be iterated regularly and migrated at any time, improving the scalability of the model and at the same time improving the efficiency of credit-granting processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings. In the drawings:

[0022] Figure 1 is a schematic flowchart of the enterprise credit-granting processing method in the embodiment of the present invention;

[0023] Figure 2 is a specific example diagram of the enterprise credit-granting processing method in the embodiment of the present invention;

[0024] Figure 3 is another specific example diagram of the enterprise credit-granting processing method in the embodiment of the present invention;

[0025] Figure 4 It is a schematic diagram of the modeling process of the first analysis module in the embodiment of the present invention;

[0026] Figure 5 It is a schematic diagram of the enterprise credit granting processing device in the embodiment of the present invention. Specific embodiments

[0027] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer and more understandable, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Herein, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but do not limit the present invention.

[0028] In the technical solution of the present application, the acquisition, transmission, storage, use, processing, etc. of data all comply with the relevant provisions of national laws and regulations.

[0029] It should be noted that in the embodiments of the present application, some existing solutions in the industry such as certain software, components, models, etc. may be mentioned. They should be regarded as exemplary. The purpose is only to illustrate the feasibility in the implementation of the technical solution of the present application, but it does not mean that the applicant has already or necessarily used this solution.

[0030] The existing enterprise credit granting processing methods in the prior art have the following disadvantages:

[0031] First, the optional features are limited. When traditional methods conduct credit granting processing, they usually select basic indicators based on customer characteristics, data availability, and expert experience. However, due to their scale, evaluation complexity, and index diversity, large and medium-sized technology-based enterprises lack a comprehensive and effective feature system when constructing business evaluation and credit analysis models, making it difficult to accurately evaluate target customers.

[0032] Second, the model scalability is weak. The current method has a long measurement process and poor scalability. Large and medium-sized technology-based enterprises are widely distributed in industries and have rapid technological iterations. The existing technology has limited flexibility and many application limitations, and cannot meet the high-intensity model expansion requirements.

[0033] Third, the maintenance cost is relatively high. When the existing technology conducts credit granting processing, it involves multiple small business modules, resulting in greater pressure on operation and maintenance costs, labor costs, monitoring costs, etc. Technology-based enterprises have rapid business changes and diverse financing methods, and have higher requirements for loan products. The high maintenance cost of the current technology cannot match the actual financial service needs of large and medium-sized technology enterprises.

[0034] In view of the above-mentioned shortcomings of the prior art, the embodiments of the present invention focus on the innovation development laws and characteristics of large and medium-sized technology-based enterprises, and use big data analysis, machine learning, and artificial intelligence technologies to establish a business evaluation and intelligent credit granting processing method for large and medium-sized technology-based enterprises by integrating the four dimensions of "looking at history, looking at capabilities, looking at demands, and looking at potential". The embodiments of the present invention can achieve the following three objectives:

[0035] First, integrate multiple internal and external data to form an underlying data system centered on financial, technological, and industrial-financial indicators, providing a comprehensive data perspective for subsequent business evaluation and the construction of credit analysis models.

[0036] Second, construct a unified business evaluation and credit granting processing methodology for large and medium-sized technology enterprises. Based on a large amount of historical approval data, mine and summarize credit granting rules and approval preferences, and launch automated business access and credit analysis models to significantly improve the quality and efficiency of credit business handling.

[0037] Third, solidify the model iteration and rule refinement process. Through constructing a scientific and technological innovation data base, a process-based feature engineering process, and solidifying the model development process, etc., realize the regular monitoring, iteration, and rule refinement of the model, and reduce the cost of manual intervention.

[0038] Figure 1 It is a schematic flow chart of the enterprise credit granting processing method in the embodiments of the present invention. As Figure 1 shown, the method includes:

[0039] Step 101, obtain the comprehensive data of the enterprise to be tested; the enterprise to be tested belongs to a large-scale enterprise or a medium-scale enterprise divided according to preset indicators; the comprehensive data includes financial data, industrial chain data, and scientific and technological innovation ability data;

[0040] Step 102, input the comprehensive data of the enterprise to be tested into a pre-constructed access model, and output the result of whether to access; the access model is pre-trained on a machine learning model using the comprehensive data of historical enterprises that have been accessed.

[0041] Step 103, when the result is access, input the comprehensive data of the enterprise to be tested into a pre-constructed credit analysis model, and output the credit risk prediction value of the enterprise to be tested; the credit analysis model includes a first analysis module, a second analysis module, and a third analysis module. The first analysis module is used to predict the financial risks of the enterprise based on an extreme gradient boosting regression model, the second analysis module is used to calculate the enterprise's contract performance ability value, and the third analysis module is used to calculate the enterprise's operation turnover ability value.

[0042] Step 104: Adjust the predicted credit risk value of the enterprise to be measured by using the science and technology innovation financial association data of the enterprise to be measured in the target bank, and output the credit evaluation result of the target bank for the enterprise to be measured; the science and technology innovation financial association data includes the information data of the existing cooperative banks of the enterprise to be measured, the science and technology innovation evaluation data of the enterprise to be measured in the target bank, and the credit data.

[0043] From Figure 1 As can be seen from the process shown, in the embodiment of the present invention, the credit processing of large and medium-sized technology-based enterprises is carried out through artificial intelligence technology. During the processing process, there are no multiple business sub-modules involved. Instead, the prediction and judgment are directly carried out through two intelligent analysis models, which releases manpower and reduces the dimensionality cost. The model can be iterated regularly and migrated at any time, improving the scalability of the model and at the same time improving the credit processing efficiency.

[0044] The enterprise credit processing method in the embodiment of the present invention will be explained in detail below.

[0045] In the embodiment of the present invention, the credit processing evaluation of large and medium-sized enterprises is carried out. The enterprise to be measured belongs to a large-scale enterprise or a medium-scale enterprise divided according to preset indicators; the preset indicators include, but are not limited to, one or any combination of turnover, total assets, and the number of employees.

[0046] For example, it is used to process the credit risk determination requirements for the operating turnover within 50 million yuan of medium-sized technology-based enterprises. Figure 2 As a specific example diagram of the enterprise credit processing method in the embodiment of the present invention, as Figure 2 shown, the overall idea of the embodiment of the present invention includes: an artificial verification link to form a customer group of large and medium-sized technology-based enterprises and conduct artificial verification according to set requirements; a machine automatic verification link to apply big data machine learning methods to establish a differentiated access model, improve the front-end customer screening ability, and use the access model to analyze and process whether the borrower is eligible for access; a borrower total risk prediction link to calculate the predicted credit risk value of the borrower; a credit risk adjustment link in the target bank to comprehensively determine the credit risk level in the target bank by combining the borrower's science and technology innovation evaluation level, credit rating, and the situation of existing cooperative banks, and finally determine the credit evaluation result in the target bank through the calculated predicted credit risk value and the market share ratio in the target bank. The credit evaluation result may include, but is not limited to, the risk level and the credit limit corresponding to the risk level. It should be noted that in the adjustment link, finally, it can be adjusted and confirmed by expert experience. Through the man-machine interaction interface, the credit evaluation results of multiple enterprises are provided, and at the same time, an information query interface for the credit evaluation process of the enterprise is provided, and it is adjusted and confirmed by expert experience.

[0047] During specific implementation, first obtain the comprehensive data of the enterprise to be measured, and the comprehensive data includes, but is not limited to, financial data, industrial chain data, and science and technology innovation ability data.

[0048] For example, a large amount of comprehensive data such as the working capital, non - working capital, liabilities, turnover, costs, cash flow from operating activities, supplier data, inventory data, R & D investment data, intellectual property data, and new product sales data of the enterprise to be measured is obtained.

[0049] In one embodiment, before inputting the comprehensive data of the enterprise to be measured into the pre - constructed access model, it may further include:

[0050] Based on the evaluation feature data in the pre - constructed science and technology innovation data base, pre - process the comprehensive data of the enterprise to be measured to obtain the feature data of the enterprise to be measured; the pre - processing includes but is not limited to one or any combination of data cleaning, data conversion, and feature extraction;

[0051] Inputting the comprehensive data of the enterprise to be measured into the pre - constructed access model includes: inputting the feature data of the enterprise to be measured into the pre - constructed access model;

[0052] Inputting the comprehensive data of the enterprise to be measured into the pre - constructed credit analysis model includes: inputting the feature data of the enterprise to be measured into the pre - constructed credit analysis model.

[0053] During implementation, a science and technology innovation data base is established in advance. The science and technology innovation data base includes multiple evaluation feature data at one or any combination of the financial index level, industrial chain level, and science and technology enterprise value level, such as various indicators.

[0054] For example, the science and technology innovation data base is constructed as follows:

[0055] Obtain the comprehensive data of multiple enterprises;

[0056] Based on various data processing methods, form multiple evaluation feature data in aspects such as the financial index level, industrial chain level, and science and technology enterprise value level; the data processing methods include methods such as natural language processing and cluster analysis.

[0057] Generally speaking, the science and technology innovation data base is constructed based on the industrial and commercial information of enterprise customers, industry - accumulated data, and external data obtained through other legal channels.

[0058] Financial index level: Based on standard financial data, after sorting out financial check - off relationships, ratio calculation, and trend processing, and then performing data cleaning, including missing rate transmission processing, value concentration, and index similarity, it is derived to more than 2,000 indicators to assist business personnel in quickly realizing financial analysis and trend prediction.

[0059] Industrial chain level: By reconstructing the industry system, hundreds of industrial chain nodes, dozens of industrial chains, and nearly a hundred industrial - finance indicators are formed to support the query of the industrial chain to which technology - based enterprises belong, upstream and downstream analysis, and macro, medium, and micro analysis.

[0060] In terms of the value of technology-based enterprises: A complete evaluation system consisting of nearly a hundred indicators is formed around the enterprise's R & D capabilities, patent application status, etc., providing a comprehensive technological perspective for management personnel and business marketing personnel.

[0061] In this example, based on various indicators in the science and technology innovation data base, the comprehensive data of the enterprise to be tested is pre-processed to form the characteristic data of the enterprise to be tested. Subsequently, access measurement and risk prediction are carried out based on the characteristic data of the enterprise to be tested.

[0062] In step 102, the comprehensive data of the enterprise to be tested is input into a pre-constructed access model, and the result of whether to grant access is output; the access model is pre-trained on a machine learning model using the comprehensive data of historical enterprises that have been granted access.

[0063] In one embodiment, the access model can be trained as follows:

[0064] Obtain the comprehensive data of historical enterprises that have been granted access;

[0065] Pre-process the comprehensive data of historical enterprises that have been granted access to obtain positive and negative sample data that affect access;

[0066] Use the positive and negative sample data to train a machine learning model to obtain a trained access model.

[0067] Taking the customers approved over the years as the total sample, sort out and summarize the data characteristics of bad samples, establish an access model, and output the evaluation score in the form of a score. In terms of index construction, based on the basic information and science and technology data of the enterprise, with relative financial indicators as the core, an index system of 8 dimensions is formed. Figure 3 This is another specific example diagram of the enterprise credit granting processing method in the embodiments of the present invention. As Figure 3 shown, based on the processing of the core enterprise basic information, financial data, and intellectual property data (such as the data obtained at the financial index level, industrial chain level, and technology-based enterprise value level), more than 2,000 evaluation feature data are derived and entered into the access feature library, forming an index system of 8 dimensions, specifically including short-term solvency, long-term solvency, operating leverage level, financing leverage level, operating profitability, input-output efficiency, investment profitability, and operation ability. After processing using clustering analysis, principal component analysis dimensionality reduction algorithm, variable importance, stability screening, etc., an access feature library for the business access model and the credit analysis model is formed.

[0068] In addition, by focusing on the business characteristics of technology-based enterprises, sorting out bad samples, fully identifying the core indicators that may affect the business conditions of enterprises, summarizing the default rules, and through multiple rounds of optimization discussions, a access model for the business of technology-based enterprises is established through training. By outputting the evaluation score in the form of a score, it is judged whether to grant access according to the evaluation score and the set evaluation score threshold, providing clear marketing guidance for corporate account managers to handle access.

[0069] In step 103, when the result is access, the comprehensive data of the enterprise to be tested is input into a pre-constructed credit analysis model, and a credit risk prediction value of the enterprise to be tested is output; the credit analysis model includes a first analysis module, a second analysis module, and a third analysis module. The first analysis module is used to predict the financial risks of the enterprise based on the extreme gradient boosting regression model. The second analysis module is used to calculate the enterprise contract performance ability value based on a first preset formula. The third analysis module is used to calculate the enterprise operation turnover ability value based on a second preset formula.

[0070] During implementation, to further empower the full life cycle service of large and medium-sized technology-based enterprises and build a smart financial service platform for technology-based enterprises, based on the main characteristics of technology-based enterprises, methods such as machine learning, solver, and expert rules are used, combined with a large amount of historical approval data, to construct an automated credit analysis model for technology-based enterprises.

[0071] Large and medium-sized technology enterprises have mature business models and stable incoming payments, and have perfect credit ratings, standardized financial statements, and sufficient historical data. Therefore, in the credit analysis process, starting from the enterprise's financial statements, comprehensively considering technology-related indicators, using big data analysis and machine learning methods to mine the high-frequency data characteristics of technology-based enterprises, and constructing a credit analysis model with the enterprise's financial report as the core. Starting from three dimensions: financial aspects, enterprise contract performance ability value, and enterprise operation turnover ability value, the overall business conditions and future trends of customers are judged to achieve accurate prediction and processing of enterprise credit.

[0072] In one embodiment, the first analysis module can be constructed as follows:

[0073] Obtain the comprehensive data of historical large-scale enterprises or medium-scale enterprises;

[0074] Based on the customer group portrait characteristics and data correlation analysis of large-scale enterprises or medium-scale enterprises, determine multiple input variable characteristics for predicting the financial risks of enterprises from the comprehensive data of historical large-scale enterprises or medium-scale enterprises;

[0075] Train, test, and validate multiple prospective prediction models using multiple input variable features to obtain each trained prospective prediction model; the prospective prediction model is a model that uses current information to predict future trends, and the prospective prediction model includes an extreme gradient boosting regression model.

[0076] According to the comparison results of the actual values and predicted values corresponding to each trained prospective prediction model, select the prospective prediction model with the best fitting degree to form the first analysis module; among them, the prospective prediction model with the best fitting degree is the extreme gradient boosting regression model; the best fitting degree means that the evaluation index exceeds the set threshold, and the evaluation index includes the coefficient of determination.

[0077] Figure 4 This is a schematic diagram of the modeling process of the first analysis module in the embodiment of the present invention. As Figure 4 shown, it includes variable selection and modeling processes. Variable screening mainly goes through steps such as missing value check, correlation screening, centrality screening, and variable importance screening. The initial variable library undergoes variable derivation, data transformation, data cleaning, and variable importance selection to form the final variables entering the model. Among them, variable derivation includes financial index derivation and technology index derivation, data transformation includes logarithmic transformation of scale-type indicators, data cleaning includes cleaning indicators with missing values, highly concentrated values, and highly similar values, and variable importance includes variable selection based on model importance, lasso selection, and change stability assessment. The modeling process includes sample preparation, multi-algorithm modeling, model parameter tuning, and determining the champion model. Among them, sample preparation includes preparing training samples, test samples, and validation samples. The training samples include historical approval data for the past 10 years, the test samples include historical approval data for the past 10 years, and the validation samples include cross-period data for the past 2 years; when performing multi-algorithm modeling, try multiple algorithms to build models, including random forest regression models, extreme gradient boosting XGB regression models, and robust models, and compare the model effects through the coefficient of determination R-squared; when tuning model parameters, through network search, manual secondary parameter tuning, etc., tune the model hyperparameters such as learning rate, number of layers, and number of neurons in each layer, and a robust regression algorithm can also be added.

[0078] By comprehensively sorting out and judging the financial indicator characteristics of the value and risk of the medium-sized enterprise customer group, relying on the basic big data of the enterprise's stock patents, finance, standards, awards and subsidies, technology recognition qualifications, financing history, etc.; combining the data characteristics of the customer group, business requirements, interpretability, data correlation, etc., determine the available data model indicators, and use them as the input variable wide table for constructing the prediction of the future financial risks of the enterprise; use common forward-looking prediction models including random forest models, XGB regression models, robust regression models, etc. for training, testing and verification. Select appropriate model evaluation indicators, such as mean square error, R-square value, etc., and select the XGB regression model with the best model evaluation effect as the champion model. Finally, use the trained XGB regression model, that is, the first analysis module, to predict the data model indicators obtained from the comprehensive data of the enterprise to be tested, and output the risk score value of the enterprise's finance.

[0079] In one embodiment, a first preset formula is formed by using a series of financial and non-financial indicators and related data to comprehensively evaluate and indirectly reflect the enterprise's contract performance ability. For example, filter out and preprocess the enterprise's capital flow ratio, quick ratio, asset-liability ratio, product qualification rate, customer complaint rate and other indicator data from the enterprise's comprehensive data, normalize them, and add them according to the preset weights to calculate the result value reflecting the enterprise's contract performance ability.

[0080] In one embodiment, a second preset formula is formed by using the indicator data that can reflect the enterprise's operation turnover ability to comprehensively evaluate and indirectly reflect the enterprise's operation turnover ability. For example, filter out and preprocess the enterprise's inventory turnover rate, working capital turnover rate, total asset turnover rate and other indicator data from the enterprise's comprehensive data, normalize them, and add them according to the set weights to calculate the result value reflecting the enterprise's operation turnover ability.

[0081] In one embodiment, input the comprehensive data of the enterprise to be tested into a pre-constructed credit analysis model, and output the credit risk prediction value of the enterprise to be tested, which may include: based on the comprehensive data of the enterprise to be tested, use the first analysis module, the second analysis module and the third analysis module for processing respectively to obtain the processing results corresponding to each analysis module; comprehensively analyze the processing results corresponding to all analysis modules to obtain the credit risk prediction value of the enterprise to be tested; the comprehensive analysis includes weighted average.

[0082] During implementation, weights are pre-assigned to the processing results of each analysis module, and through weighted average, the result values of the three analysis modules are added together, and the added result is the credit risk prediction value of the enterprise to be tested.

[0083] Finally, using the science and technology innovation financial association data of the enterprise to be measured in the target bank, adjust the predicted credit risk value of the enterprise to be measured, and output the credit assessment result of the target bank for the enterprise to be measured; the science and technology innovation financial association data includes the information data of the existing cooperative banks of the enterprise to be measured, the science and technology innovation evaluation data and credit data of the enterprise to be measured in the target bank; the credit assessment result includes multiple risk levels and credit limits.

[0084] For example, using the science and technology innovation financial association data of the enterprise to be measured in the target bank to adjust the predicted credit risk value of the enterprise to be measured and output the credit assessment result of the target bank for the enterprise to be measured may include:

[0085] According to the preset adjustment ratios corresponding to the information data of the existing cooperative banks of the enterprise to be measured, the science and technology innovation evaluation data and credit data of the enterprise to be measured in the target bank, adjust the predicted credit risk value of the enterprise to be measured in turn;

[0086] Take the risk level corresponding to the adjusted value of the predicted credit risk value of the enterprise to be measured as the credit assessment result of the target bank for the enterprise to be measured; wherein the corresponding relationship between the predicted credit risk value and the risk level is pre-configured.

[0087] Reference Figure 2 , from the information of the existing cooperative banks of the enterprise to be measured, the science and technology innovation evaluation level and credit level of the enterprise to be measured in the target bank, respectively or in turn adjust the calculated predicted risk value, and determine the market share ratio in the target bank, and adjust to obtain the final credit risk level of the target bank for the enterprise to be measured.

[0088] For example, it is divided into first-level risk, second-level risk, third-level risk, etc. When the result is first-level risk, it means that the credit risk of the enterprise to be measured in the target bank is relatively high, and it is not recommended to grant credit. When the result is third-level risk, it means that the credit risk of the enterprise to be measured in the target bank is relatively low, and credit can be granted.

[0089] The above three analysis modules respectively calculate the total risk situation at the customer level from the three aspects of the enterprise's finance, the enterprise's contract performance ability, and the enterprise's operation and turnover ability. When finally applied, it is adjusted purposefully in combination with the science and technology innovation financial association data of the enterprise to be measured in the target bank, realizing efficient processing of credit for all science and technology customers, especially large and medium-sized enterprises.

[0090] By integrating a large amount of internal and external data, combining the development laws of technology-based enterprises and actual business needs, deriving numerous effective features, and using big data analysis, machine learning and artificial intelligence technologies, the embodiment of the present invention constructs a business access and credit analysis model suitable for all technology-based enterprises, helps to quickly screen and lock target customers and provides effective credit processing support, and improves the intelligent level of credit approval decision-making for technology-based enterprises.

[0091] In the embodiment of the present invention, the enterprise credit granting processing method is an information processing method in which all steps are implemented by devices such as computers.

[0092] In one embodiment, the second analysis module in the embodiment of the present invention can also predict the enterprise contract performance ability value through a decision tree model. For example, based on the comprehensive data of multiple large and medium-sized enterprises in the science and technology innovation data base, data preprocessing and feature selection are performed to obtain multiple appropriate feature data, and the data set is recursively divided according to indicators such as the information gain of the features to construct a decision tree.

[0093] For example, taking the employee turnover rate as a node for division, if the turnover rate is higher than a certain threshold, other features such as the management level are further examined, and finally a decision tree structure capable of judging the risk level of the enterprise contract performance ability is formed. During implementation, the training set data is used to train the model, and model parameters such as the maximum depth of the tree and the minimum number of samples for splitting are adjusted to optimize the model performance, forming a trained second analysis module. Finally, the trained second analysis module is used to predict the feature data label obtained based on the comprehensive data of the enterprise to be tested, and the risk score value in terms of the enterprise contract performance ability is output.

[0094] In one embodiment, the third analysis module in the embodiment of the present invention can also predict the enterprise operation turnover ability value through a regression model. For example, based on the comprehensive data of multiple large and medium-sized enterprises in the science and technology innovation data base, a regression model between the risk score in terms of the enterprise operation turnover ability and various influencing factors of the operation turnover ability is established, and the influence degree and relationship of each factor on the enterprise operation turnover ability are determined by analyzing historical data, so as to predict the risk score value in terms of the operation turnover ability of the enterprise to be tested. The enterprise operation turnover ability factors are, for example, factors such as inventory management, working capital, production organization efficiency, and supply chain management level.

[0095] Finally, when fusing the processing results of the three analysis modules, the Bayesian method can also be used for fusion. During implementation, first, the three analysis modules are used to complete the calculation, and then the core financial indicators are used to classify the customer group to determine the customer group classification of the enterprise to be tested. Different customer groups use different rules to fuse the prediction results of the three analysis modules. For example, the customer groups include customer groups with different industry prospects, different development trends, and different degrees of talent and resource attraction. For example, for an enterprise to be tested with a very good industry prospect (quantifying the industry prospect development trend value, when the industry prospect development trend value exceeds the first preset value, it is a very good industry prospect), the Bayesian method is used for fusion, and for an enterprise to be tested with an average industry prospect (when the industry prospect development trend value is lower than the second preset value, it is an average industry prospect), the weighted average method is used for fusion.

[0096] An enterprise credit granting processing device is also provided in an embodiment of the present invention, as described in the following embodiments. Since the principle of the device for solving problems is similar to that of the enterprise credit granting processing method, the implementation of the device can refer to the implementation of the enterprise credit granting processing method, and the repeated parts will not be elaborated.

[0097] Figure 5 It is a schematic diagram of the enterprise credit granting processing device in an embodiment of the present invention. As Figure 5 shown, the device 500 includes:

[0098] A data acquisition module 501, configured to acquire comprehensive data of an enterprise to be measured; the enterprise to be measured belongs to a large-scale enterprise or a medium-scale enterprise divided according to preset indicators; the comprehensive data includes financial data, industrial chain data, and scientific and technological innovation ability data;

[0099] An access judgment module 502, configured to input the comprehensive data of the enterprise to be measured into a pre-constructed access model, and output a result of whether to grant access; the access model is pre-trained by using the comprehensive data of historical enterprises that have been granted access to a machine learning model;

[0100] A credit analysis module 503, configured to, when the result is access, input the comprehensive data of the enterprise to be measured into a pre-constructed credit analysis model, and output a credit risk prediction value of the enterprise to be measured; the credit analysis model includes a first analysis module, a second analysis module, and a third analysis module. The first analysis module is used to predict the risk in the financial aspect of the enterprise based on an extreme gradient boosting regression model, the second analysis module is used to calculate the enterprise contract performance ability value, and the third analysis module is used to calculate the enterprise operation turnover ability value;

[0101] An adjustment module 504, configured to adjust the credit risk prediction value of the enterprise to be measured by using the scientific and technological innovation financial association data of the enterprise to be measured in the target bank, and output a credit evaluation result of the target bank for the enterprise to be measured; the scientific and technological innovation financial association data includes the information data of the existing cooperative banks of the enterprise to be measured, the scientific and technological innovation evaluation data and credit data of the enterprise to be measured in the target bank.

[0102] In one embodiment, the preset indicators include one or any combination of turnover, total assets, and the number of employees.

[0103] In one embodiment, the device 500 further includes:

[0104] A preprocessing module, configured to preprocess the comprehensive data of the enterprise to be measured based on the evaluation feature data in a pre-constructed scientific and technological innovation data base to obtain the feature data of the enterprise to be measured before the access judgment module 502 inputs the comprehensive data of the enterprise to be measured into a pre-constructed access model; the preprocessing includes one or any combination of data cleaning, data conversion, and feature extraction;

[0105] The access judgment module 502 is specifically configured to: input the characteristic data of the enterprise to be measured into a pre-constructed access model;

[0106] The credit analysis module 503 is specifically configured to: input the characteristic data of the enterprise to be measured into a pre-constructed credit analysis model.

[0107] In one embodiment, the science and technology innovation data base includes multiple evaluation characteristic data at one or any combination of the financial index level, the industrial chain level, and the science and technology enterprise value level.

[0108] In one embodiment, the access model is trained as follows:

[0109] Obtain the comprehensive data of historical enterprises that have been granted access;

[0110] Preprocess the comprehensive data of historical enterprises that have been granted access to obtain positive and negative sample data that affect access;

[0111] Use the positive and negative sample data to train a machine learning model to obtain a trained access model.

[0112] In one embodiment, the first analysis module is constructed as follows:

[0113] Obtain the comprehensive data of historical large-scale or medium-scale enterprises;

[0114] Based on the customer group portrait characteristics and data correlation analysis of large-scale or medium-scale enterprises, determine multiple input variable characteristics for predicting the financial risks of enterprises from the comprehensive data of historical large-scale or medium-scale enterprises;

[0115] Use multiple input variable characteristics to train, test, and verify multiple forward-looking prediction models to obtain each trained forward-looking prediction model; the forward-looking prediction model is a model that uses current information to predict future trends, and the forward-looking prediction model includes an extreme gradient boosting regression model;

[0116] According to the comparison results of the actual values and predicted values corresponding to each trained forward-looking prediction model, select the forward-looking prediction model with the best fitting degree to form the first analysis module; the forward-looking prediction model with the best fitting degree is an extreme gradient boosting regression model; the best fitting degree means that the evaluation index exceeds the set threshold, and the evaluation index includes the coefficient of determination.

[0117] In one embodiment, the credit analysis module 503 is specifically configured to:

[0118] Based on the comprehensive data of the enterprise to be measured, the first analysis module, the second analysis module, and the third analysis module are respectively used for processing to obtain the processing results corresponding to each analysis module;

[0119] The processing results corresponding to all analysis modules are comprehensively analyzed to obtain the credit risk prediction value of the enterprise to be measured; the comprehensive analysis includes weighted average.

[0120] In one embodiment, the adjustment module 504 is specifically configured to:

[0121] According to the preset adjustment ratios corresponding to the existing cooperative bank information data of the enterprise to be measured, the science and technology innovation evaluation data and credit data of the enterprise to be measured in the target bank, the credit risk prediction value of the enterprise to be measured is adjusted in sequence;

[0122] The risk level corresponding to the adjusted value of the credit risk prediction value of the enterprise to be measured is used as the credit evaluation result of the target bank for the enterprise to be measured; wherein the corresponding relationship between the credit risk prediction value and the risk level is pre-configured.

[0123] An embodiment of the present invention further provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the above-mentioned enterprise credit processing method is implemented.

[0124] An embodiment of the present invention further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned enterprise credit processing method is implemented.

[0125] An embodiment of the present invention further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the above-mentioned enterprise credit processing method is implemented.

[0126] In the embodiments of the present invention, for large-scale enterprises or medium-scale enterprises, comprehensive data is obtained. First, a machine learning model is used to determine whether a to-be-tested enterprise is eligible for access. When the result is eligible for access, a pre-constructed credit analysis model is used to predict the credit risk prediction value of the to-be-tested enterprise. The credit analysis model includes a first analysis module for predicting the financial risks of the enterprise, a second analysis module for calculating the enterprise's contract performance ability value, and a third analysis module for calculating the enterprise's operation turnover ability value. Finally, using the science and technology innovation financial association data of the to-be-tested enterprise in the target bank, the credit risk prediction value of the to-be-tested enterprise is adjusted, and the credit assessment result of the target bank for the to-be-tested enterprise is output. Compared with the prior art, the embodiments of the present invention perform credit analysis and processing on large and medium-sized technology-based enterprises through artificial intelligence technology. During the processing, multiple small business modules are not involved, and prediction and judgment are directly carried out through two intelligent analysis models, which releases manpower and reduces dimensionality costs. The model can be iterated regularly and migrated at any time, improving the scalability of the model and at the same time improving the credit processing efficiency.

[0127] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0128] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for realizing the functions specified in Figure 1 one or more flows or multiple flows and / or blocks Figure 1 one or more blocks or multiple blocks.

[0129] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device realizes the functions specified in Figure 1 one or more flows or multiple flows and / or blocks Figure 1 one or more blocks or multiple blocks.

[0130] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are executed on the computer or other programmable apparatus to generate a computer-implemented process, thereby providing instructions for implementing the steps of the function specified in one process or a plurality of processes and / or one block or a plurality of blocks in the flow Figure 1 One process or a plurality of processes and / or Figure 1 steps of the function specified in one block or a plurality of blocks.

[0131] The specific embodiments described above further elaborate on the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for processing corporate credit, characterized in that: include: Obtain comprehensive data of the enterprise to be tested; The enterprise to be tested belongs to a large-scale enterprise or a medium-sized enterprise classified according to preset indicators; the comprehensive data includes financial data, industrial chain data, and scientific and technological innovation capability data; Input the comprehensive data of the enterprise to be tested into the pre-built admission model and output the result of whether to allow admission; The admission model is obtained by pre-training a machine learning model using the comprehensive data of historically admitted enterprises; When the result is admission, the comprehensive data of the enterprise to be tested is input into the pre-built credit analysis model, and the credit risk prediction value of the enterprise to be tested is output; The credit analysis model includes a first analysis module, a second analysis module and a third analysis module, wherein the first analysis module is used to predict the financial risk of the enterprise based on the extreme gradient boosting regression model, the second analysis module is used to calculate the contract performance capacity value of the enterprise, and the third analysis module is used to calculate the operation turnover capacity value of the enterprise; The scientific and technological innovation finance-related data of the enterprise to be tested in the target bank are used to adjust the credit risk prediction value of the enterprise to be tested, and the credit assessment result of the target bank for the enterprise to be tested is output; the scientific and technological innovation finance-related data include the existing cooperative bank information data of the enterprise to be tested, the scientific and technological innovation evaluation data and credit data of the enterprise to be tested in the target bank.

2. The method according to claim 1, characterized in that The preset indicators include turnover, total assets, number of employees, or any combination thereof.

3. The method according to claim 1, characterized in that Before the comprehensive data of the companies to be tested is input into the pre-built admission model, it also includes: Based on the evaluation feature data in the pre-built science and technology innovation data base, the comprehensive data of the enterprise to be tested is pre-processed to obtain the feature data of the enterprise to be tested; the pre-processing includes one or any combination of data cleaning, data conversion, and feature extraction; Comprehensive data on the companies to be tested is fed into a pre-built admission model, including: Input the characteristic data of the enterprise to be tested into the pre-built admission model; Input the comprehensive data of the enterprise to be tested into the pre-built credit analysis model, including: Input the characteristic data of the enterprise to be tested into the pre-built credit analysis model.

4. The method according to claim 3, characterized in that The scientific and technological innovation data base includes multiple evaluation feature data at one or any combination of the financial indicator level, the industrial chain level, and the scientific and technological enterprise value level.

5. The method according to claim 1, characterized in that The admission model is trained as follows: Obtain comprehensive data on historically admitted enterprises; Pre-process the comprehensive data of historical enterprises that have been admitted to obtain positive and negative sample data that have an impact on admission; Use positive and negative sample data to train the machine learning model to obtain a trained access model.

6. The method according to claim 1, characterized in that The first analysis module is constructed as follows: Obtain comprehensive data on historical large-scale or medium-sized enterprises; Based on the customer profile characteristics and data correlation analysis of large-scale or medium-sized enterprises, multiple input variable characteristics for predicting corporate financial risks are determined from the historical comprehensive data of large-scale or medium-sized enterprises; Using multiple input variable features, multiple forward-looking prediction models are trained, tested and verified to obtain each trained forward-looking prediction model; the forward-looking prediction model is a model that uses current information to predict future trends, and the forward-looking prediction model includes an extreme gradient boosting regression model; According to the comparison results of the actual values ​​and predicted values ​​corresponding to each trained forward-looking prediction model, the forward-looking prediction model with the best fitting degree is selected to form the first analysis module; wherein the forward-looking prediction model with the best fitting degree is the extreme gradient boosting regression model; the best fitting degree indicates that the evaluation index exceeds the set threshold, and the evaluation index includes the determination coefficient.

7. The method according to claim 1, characterized in that Input the comprehensive data of the enterprise to be tested into the pre-built credit analysis model, and output the credit risk prediction value of the enterprise to be tested, including: Based on the comprehensive data of the enterprise to be tested, the first analysis module, the second analysis module and the third analysis module are used for processing respectively to obtain the processing results corresponding to each analysis module; Comprehensively analyze the processing results corresponding to all analysis modules to obtain the credit risk prediction value of the enterprise to be tested; the comprehensive analysis includes a weighted average method.

8. The method according to claim 1, characterized in that The credit assessment result includes the risk level; using the science and technology innovation finance related data of the enterprise to be tested in the target bank, the credit risk prediction value of the enterprise to be tested is adjusted, and the credit assessment result of the target bank for the enterprise to be tested is output, including: According to the preset adjustment ratios corresponding to the existing cooperative bank information data of the enterprise to be tested, the scientific and technological innovation evaluation data and credit data of the enterprise to be tested in the target bank, the credit risk prediction value of the enterprise to be tested is adjusted in sequence; The risk level corresponding to the adjusted value of the credit risk prediction value of the enterprise to be tested is used as the credit assessment result of the target bank for the enterprise to be tested; wherein the corresponding relationship between the credit risk prediction value and the risk level is pre-configured.

9. An enterprise credit processing device, characterized in that: include: Data acquisition module, used to obtain comprehensive data of the enterprise to be tested; The enterprise to be tested belongs to a large-scale enterprise or a medium-sized enterprise classified according to preset indicators; the comprehensive data includes financial data, industrial chain data, and scientific and technological innovation capability data; The admission judgment module is used to input the comprehensive data of the enterprise to be tested into a pre-built admission model and output the result of whether to admit or not; the admission model is pre-trained by using the comprehensive data of the historical enterprises that have been admitted to train the machine learning model; The credit analysis module is used to input the comprehensive data of the enterprise to be tested into the pre-built credit analysis model when the result is admission, and output the credit risk prediction value of the enterprise to be tested; The credit analysis model includes a first analysis module, a second analysis module and a third analysis module, wherein the first analysis module is used to predict the financial risk of the enterprise based on the extreme gradient boosting regression model, the second analysis module is used to calculate the contract performance capacity value of the enterprise, and the third analysis module is used to calculate the operation turnover capacity value of the enterprise; The adjustment module is used to adjust the credit risk prediction value of the enterprise to be tested by using the science and technology innovation finance related data of the enterprise to be tested in the target bank, and output the credit assessment result of the target bank for the enterprise to be tested; the science and technology innovation finance related data includes the existing cooperative bank information data of the enterprise to be tested, the science and technology innovation evaluation data and credit data of the enterprise to be tested in the target bank.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 8 is implemented.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

12. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images