Enterprise credit line risk assessment method and system

Abstract

The invention discloses an enterprise credit line risk assessment method and system, and the method comprises the steps: collecting enterprise information corresponding to a to-be-assessed enterprise, the enterprise information including but not limited to enterprise basic information, an enterprise financial report, an enterprise tax record, industry dynamic information, and financing investment information; the enterprise information is preprocessed, noise information and abnormal information in the enterprise information are removed, and to-be-evaluated information is obtained; according to a preset credit granting keyword, extracting credit granting associated data in the to-be-evaluated information; and according to an industry evaluation module corresponding to the to-be-evaluated enterprise in a preset neural network model, analyzing and evaluating the credit granting keyword and the credit granting associated data to obtain a credit coefficient of the to-be-evaluated enterprise, and according to the credit coefficient, evaluating a credit line risk of the to-be-evaluated enterprise. According to the invention, comprehensive and accurate assessment of the credit line risk of the enterprise is realized.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method and system for assessing enterprise credit line risk. Background Art

[0002] The credit line of an enterprise is a loan amount granted by a bank or other financial institution to an enterprise for use within a certain period of time based on the enterprise's credit status, operating ability, financial status, etc. In order to prevent risks and ensure that the enterprise can repay the loan in full and on time, banks or other financial institutions usually need to assess the credit line risk of the enterprise.

[0003] At present, the risk of corporate credit lines is mainly assessed by static data such as financial statements and historical transaction records. On the one hand, although such data has certain reference value, it is still difficult to fully reflect the current operating status and potential risks of the enterprise. On the other hand, it is easy to lead to inaccurate assessment due to insufficient in-depth assessment and analysis. Therefore, how to comprehensively and accurately assess the risk of corporate credit lines is one of the technical problems that need to be solved at present. Summary of the invention

[0004] The main purpose of the present invention is to provide a method and system for enterprise credit line risk assessment, aiming to solve the technical problem of how to comprehensively and accurately conduct enterprise credit line risk assessment in the prior art.

[0005] To achieve the above object, the present invention provides a method for assessing the risk of an enterprise credit line, the method comprising:

[0006] Collect enterprise information corresponding to the enterprise to be evaluated, including but not limited to basic enterprise information, enterprise financial reports, enterprise tax records, industry dynamics information, and financing and investment information;

[0007] Preprocessing the enterprise information to remove noise information and abnormal information in the enterprise information to obtain information to be evaluated;

[0008] Extracting credit-related data from the information to be evaluated according to preset credit keywords;

[0009] According to the industry assessment module corresponding to the enterprise to be assessed in the preset neural network model, the credit keywords and credit-related data are analyzed and evaluated to obtain the credit coefficient of the enterprise to be assessed, and the credit line risk of the enterprise to be assessed is evaluated based on the credit coefficient.

[0010] Preferably, the step of evaluating the credit line risk of the enterprise to be evaluated according to the credit coefficient includes:

[0011] Finding a credit score and a risk level corresponding to the credit coefficient, and obtaining first credit limit information, first interest rate information, and first guarantee information corresponding to the credit score, and second credit limit information, second interest rate information, and second guarantee information corresponding to the risk level;

[0012] Finding the credit weight and risk weight corresponding to the credit score and risk level, and weighting the first credit limit information, the first interest rate information, the first guarantee information, the second credit limit information, the second interest rate information, and the second guarantee information according to the credit weight and risk weight to obtain credit limit weight information, interest rate weight information, and guarantee weight information;

[0013] The credit score, risk level, credit line weight information, interest rate weight information and guarantee weight information are jointly generated as credit reference information and output, thereby completing the assessment of the credit line risk of the enterprise to be assessed.

[0014] Preferably, the step of extracting the credit-related data in the information to be evaluated according to the preset credit keywords includes:

[0015] Determine the number of single Chinese characters contained in the preset credit keyword, and use each character in the data to be evaluated as the first character, perform word segmentation processing according to the number, and obtain multiple keywords;

[0016] Calculate the similarity values ​​between the preset credit keyword and the multiple keywords respectively, and compare the similarity values ​​to determine the maximum value among the similarity values, wherein the similarity value calculation formula is:

[0017]

[0018] Where S represents the similarity value, W represents the word vector of the preset credit keyword, and Wi represents the word vector of the i-th keyword;

[0019] A target keyword corresponding to the maximum value is searched in each of the keywords, and a value corresponding to the target keyword is searched in the information to be evaluated as the credit-related data.

[0020] Preferably, the step of analyzing and evaluating the credit keywords and credit-related data according to the industry evaluation module corresponding to the enterprise to be evaluated in the preset neural network model comprises:

[0021] Collecting credit training samples corresponding to the enterprise to be evaluated, and training the initial evaluation module corresponding to the enterprise to be evaluated in the preset initial network model according to the credit training samples;

[0022] When the training duration reaches the preset duration, the loss function value corresponding to the initial evaluation module is calculated, and it is determined whether the calculated loss function value is less than the preset loss threshold, wherein the calculation formula of the loss function value is:

[0023]

[0024] Wherein, L represents the loss function value, n represents the number of credit training samples, yi represents the training value obtained by training the i-th credit training sample, and yi0 represents the reference value corresponding to the i-th credit training sample;

[0025] If it is less than the preset loss threshold, the training of the initial assessment module is terminated, and the initial assessment module is generated as the industry assessment module;

[0026] If the loss function value is not less than the preset loss threshold, executing the step of training the initial assessment module corresponding to the enterprise to be assessed in the preset initial network model according to the credit training sample until the loss function value is less than the preset loss threshold;

[0027] When all modules included in the preset initial network model have completed training, the preset initial network model is generated as the preset neural network model.

[0028] Preferably, the step of evaluating the credit line risk of the enterprise to be evaluated according to the credit coefficient includes:

[0029] According to the credit keyword and the credit associated data, the credit training sample is updated, and an update rate of the credit training sample is generated;

[0030] Determine whether the update rate is greater than a preset update threshold. If so, perform a step of training an initial assessment module corresponding to the enterprise to be assessed in a preset initial network model based on the updated credit training sample.

[0031] Furthermore, to achieve the above-mentioned purpose, the present invention also provides a corporate credit line risk assessment system, the corporate credit line risk assessment system comprising:

[0032] A collection module is used to collect enterprise information corresponding to the enterprise to be evaluated, including but not limited to basic enterprise information, enterprise financial reports, enterprise tax records, industry dynamic information, and financing and investment information;

[0033] A preprocessing module, used to preprocess the enterprise information, remove noise information and abnormal information in the enterprise information, and obtain information to be evaluated;

[0034] An extraction module, used to extract the credit-related data in the information to be evaluated according to preset credit keywords;

[0035] An evaluation module is used to analyze and evaluate the credit keywords and credit-related data according to the industry evaluation module corresponding to the enterprise to be evaluated in the preset neural network model, obtain the credit coefficient of the enterprise to be evaluated, and evaluate the credit line risk of the enterprise to be evaluated based on the credit coefficient.

[0036] Preferably, the evaluation module further includes:

[0037] an acquisition unit, configured to search for a credit score and a risk level corresponding to the credit coefficient, and acquire first credit limit information, first interest rate information, and first guarantee information corresponding to the credit score, and second credit limit information, second interest rate information, and second guarantee information corresponding to the risk level;

[0038] a weighting unit, configured to search for a credit weight and a risk weight corresponding to the credit score and the risk level, and perform weighted processing on the first credit limit information, the first interest rate information, the first guarantee information, the second credit limit information, the second interest rate information, and the second guarantee information according to the credit weight and the risk weight, so as to obtain credit limit weight information, interest rate weight information, and guarantee weight information;

[0039] The generating unit is used to generate the credit score, risk level, credit limit weight information, interest rate weight information and guarantee weight information into credit reference information and output it, so as to complete the credit limit risk assessment of the enterprise to be assessed.

[0040] Preferably, the extraction module comprises:

[0041] A determination unit, used to determine the number of single Chinese characters contained in the preset credit keyword, and use each character in the data to be evaluated as the first character, perform word segmentation processing according to the number, and obtain multiple keywords;

[0042] The comparison unit is used to calculate the similarity values ​​between the preset credit keyword and the plurality of keywords respectively, and compare the similarity values ​​to determine the maximum value among the similarity values, wherein the similarity value calculation formula is:

[0043]

[0044] Where S represents the similarity value, W represents the word vector of the preset credit keyword, and Wi represents the word vector of the i-th keyword;

[0045] A searching unit is used to search for a target keyword corresponding to the maximum value in each of the keywords, and to search for a value corresponding to the target keyword in the information to be evaluated as the credit-related data.

[0046] Preferably, the enterprise credit line risk assessment system further includes:

[0047] A training module, used for collecting credit training samples corresponding to the enterprise to be evaluated, and training the initial evaluation module corresponding to the enterprise to be evaluated in the preset initial network model according to the credit training samples;

[0048] The calculation module is used to calculate the loss function value corresponding to the initial evaluation module when the training duration reaches the preset duration, and determine whether the calculated loss function value is less than the preset loss threshold, wherein the calculation formula of the loss function value is:

[0049]

[0050] Wherein, L represents the loss function value, n represents the number of credit training samples, yi represents the training value obtained by training the i-th credit training sample, and yi0 represents the reference value corresponding to the i-th credit training sample;

[0051] A generation module, when the loss function value is less than a preset loss threshold, ends the training of the initial evaluation module and generates the initial evaluation module as the industry evaluation module;

[0052] An execution module, configured to execute, when the loss function value is not less than a preset loss threshold, a step of training an initial evaluation module corresponding to the enterprise to be evaluated in a preset initial network model according to the credit training sample, until the loss function value is less than the preset loss threshold;

[0053] The generating module is further used to generate the preset initial network model into the preset neural network model when all modules included in the preset initial network model have completed training.

[0054] Preferably, the enterprise credit line risk assessment system further includes:

[0055] An updating module, used for updating the credit training sample according to the credit keyword and the credit associated data, and generating an update rate of the credit training sample;

[0056] The judgment module is used to judge whether the update rate is greater than a preset update threshold. If it is greater than the preset update threshold, the step of training the initial evaluation module corresponding to the enterprise to be evaluated in the preset initial network model according to the credit training sample is executed based on the updated credit training sample.

[0057] The enterprise credit line risk assessment method and system of the present invention collects enterprise information corresponding to the enterprise to be assessed, and the collected enterprise information at least includes enterprise basic information, enterprise financial report, enterprise tax record, industry dynamic information, and financing investment information; and pre-processes the collected enterprise information to remove noise information and abnormal information therein to obtain the information to be assessed; and extracts credit-related data in the information to be assessed according to preset credit keywords; and then analyzes and evaluates the credit keywords and credit-related data according to the industry assessment module corresponding to the enterprise to be assessed in the preset neural network model, obtains the credit coefficient of the enterprise to be assessed, and evaluates the credit line risk of the enterprise to be assessed according to the credit coefficient. In this way, various types of information related to the enterprise are collected with the help of crawler technology, so that the collected information is more comprehensive, and then the credit-related data obtained after cleaning, screening and extraction of various types of information are analyzed and evaluated by the preset neural network model formed by training a large amount of sample data, so that the generated credit coefficient for assessing the enterprise credit line risk is more accurate, and a comprehensive and accurate assessment of the enterprise credit line risk is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 This is a module diagram of the first embodiment of the enterprise credit line risk assessment method of the present invention;

[0059] Figure 2 This is a module diagram of the second embodiment of the enterprise credit line risk assessment method of the present invention;

[0060] Figure 3 This is a flow chart of the first embodiment of the enterprise credit line risk assessment system of the present invention;

[0061] Figure 4 It is a flow chart of the second embodiment of the enterprise credit line risk assessment system of the present invention.

[0062] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0063] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0064] The present invention provides a method for assessing the risk of corporate credit line. Figure 1 , Figure 1The flowchart of the first embodiment of the enterprise credit line risk assessment method of the present invention is shown in the flowchart. It should be noted that although the logical order is shown in the flowchart, in some cases, the steps shown or described can be performed in a different order than here. Specifically, the enterprise credit line risk assessment method in this embodiment includes:

[0065] Step S10, collecting enterprise information corresponding to the enterprise to be evaluated, the enterprise information includes but is not limited to basic enterprise information, enterprise financial reports, enterprise tax records, industry dynamic information, and financing and investment information;

[0066] The enterprise credit line risk assessment method of this embodiment is applied to the enterprise credit line risk assessment system. When an enterprise needs to initiate a financial loan from a bank or other financial institution, the bank or other financial institution initiates an assessment request to the enterprise credit line risk assessment system. After receiving the assessment request, the enterprise credit line risk assessment system reads the enterprise ID (Identification) carried therein, determines the specific enterprise that needs to undergo credit information assessment based on the enterprise ID, and uses the enterprise that needs to undergo credit information assessment as the enterprise to be assessed.

[0067] Furthermore, the enterprise credit line risk assessment system starts the enterprise credit line risk assessment for the enterprise to be assessed, and first collects various enterprise information related to the enterprise to be assessed from various official websites, news websites, social platforms and other channels by calling API, web crawlers and other technologies. Among them, according to the nature of the website where the enterprise information comes from and the category of the information itself, the collected enterprise information at least includes basic enterprise information indicating the nature and scale of the enterprise, enterprise financial reports, enterprise tax information, industry dynamic information and financing investment information.

[0068] Step S20, preprocessing the enterprise information, removing noise information and abnormal information in the enterprise information, and obtaining information to be evaluated;

[0069] Understandably, the accuracy of various enterprise information collected varies, so the collected enterprise information needs to be preprocessed. Preprocessing includes data cleaning, eliminating repeated noise information that is irrelevant to the enterprise to be evaluated, and screening out abnormal information. Among them, the screening of abnormal information can be based on the consistency between the various information in the collected enterprise information. This consistency can be represented by the contradiction between the same type of data from different websites. If there is a contradiction between the same type of data, it means that the consistency between the data is low, and accurate data needs to be screened out. If there is no contradiction between the various types of data, it means that the consistency between the data is high and the data is relatively accurate. The order of screening is official data first, that is, official data shall prevail. After eliminating data irrelevant to the enterprise, removing duplicates, and screening out accurate data, the information to be evaluated related to the enterprise to be evaluated is obtained.

[0070] Step S30, extracting the credit-related data in the information to be evaluated according to the preset credit keywords;

[0071] Furthermore, banks and corporate financial institutions can set keywords related to credit line risks that they are concerned about as preset credit keywords based on their previous experience in corporate credit line risk assessment. The corporate credit line risk assessment system performs text recognition on the information to be assessed based on the preset credit keywords, and extracts text related to the corporate credit line in the information to be assessed through text recognition as credit-related data corresponding to the enterprise to be assessed. Specifically, the step of extracting credit-related data from the information to be assessed based on the preset credit keywords includes:

[0072] Step S31, determining the number of single Chinese characters contained in the preset credit keyword, and taking each character in the data to be evaluated as the first character, performing word segmentation processing according to the number, and obtaining multiple keywords;

[0073] Step S32, calculating similarity values ​​between the preset credit keyword and the plurality of keywords respectively, and comparing the similarity values ​​to determine the maximum value among the similarity values;

[0074] Step S33, searching for a target keyword corresponding to the maximum value in each of the keywords, and searching for a value corresponding to the target keyword in the information to be evaluated as the credit-related data.

[0075] Further, count the number of individual Chinese characters contained in the preset credit-granting keywords, and perform word segmentation on the data to be evaluated according to this number. After removing the numbers from the data to be evaluated, take each character as the first character and perform word segmentation according to the number to obtain multiple keywords. For example, if the counted number is 3 and the data to be evaluated is "In the first half of 2024, the company achieved a total revenue of RMB A", then take "上", "半", "年", "公", "司", "实", "现", "了", "总", "收", "入", "人", "民", "币" as the first characters respectively, and perform word segmentation according to the number 3 to obtain multiple keywords such as "上半年", "半年公", "年公司"... Among them, in order to perform more accurate word segmentation, before word segmentation, the modal particles and conjunctions in the data to be evaluated can be removed. For example, the "了" in the above data to be evaluated. After removal, perform word segmentation with each character as the first character to obtain multiple more accurate keywords.

[0076] Further, a vector dictionary and a calculation formula for calculating the word segmentation similarity are preset in the enterprise credit limit risk assessment system. Search for the preset credit-granting keywords and the word vectors corresponding to each keyword from the vector dictionary, and calculate the similarity values between the word vectors of the preset credit-granting keywords and the word vectors of each keyword through the calculation formula. The specific calculation formula can be seen in the following formula (1).

[0077]

[0078] Among them, S represents the similarity value, W represents the word vector of the preset credit-granting keyword, Wi represents the word vector of the i-th keyword, W·Wi is the vector dot product operation, and ‖W‖ and ‖Wi‖ are the vector modulus operations.

[0079] Further, compare each similarity value to determine the maximum value among them. Then search for the keyword that generates this maximum value among each keyword as the target keyword, and search for the value corresponding to the target keyword in the information to be evaluated as the credit-related data. For example, for the above information to be evaluated "In the first half of 2024, the company achieved a total revenue of RMB A", if the target keyword is "total revenue", then the value A is the credit-related data. Among them, multiple preset credit-granting keywords can be set. After each piece of information to be evaluated is extracted by each preset credit-granting keyword, multiple credit-related data related to the enterprise to be evaluated are obtained.

[0080] Step S40, according to the industry evaluation module corresponding to the enterprise to be evaluated in the preset neural network model, analyze and evaluate the credit-granting keywords and the credit-related data, obtain the credit coefficient of the enterprise to be evaluated, and evaluate the credit limit risk of the enterprise to be evaluated according to the credit coefficient.

[0081] Furthermore, a preset neural network model is pre-formed through training in the enterprise credit line risk assessment system. And, in order to better reflect the characteristics of enterprises in different industries, the preset neural network model can be trained according to the industry during training, so that the preset neural network model contains multiple industry assessment modules, and each industry assessment module corresponds to an industry. For the enterprise to be assessed, first determine the industry in which it is located, and then find the industry assessment module corresponding to the industry in the preset neural network model as the industry assessment module corresponding to the enterprise to be assessed. According to the industry assessment module, the credit keywords and the credit-related data corresponding to each credit keyword are analyzed and evaluated to obtain the credit coefficient of the enterprise to be assessed. The credit coefficient represents the degree of loan risk of the enterprise to be assessed, and can be used to evaluate the credit line risk of the enterprise to be assessed.

[0082] Wherein, the step of evaluating the credit line risk of the enterprise to be evaluated according to the credit coefficient includes:

[0083] Step S41, searching for a credit score and a risk level corresponding to the credit coefficient, and obtaining first credit limit information, first interest rate information, and first guarantee information corresponding to the credit score, and second credit limit information, second interest rate information, and second guarantee information corresponding to the risk level;

[0084] Step S42, searching for the credit weight and risk weight corresponding to the credit score and risk level, and weighting the first credit limit information, the first interest rate information, the first guarantee information, the second credit limit information, the second interest rate information, and the second guarantee information according to the credit weight and risk weight to obtain credit limit weight information, interest rate weight information, and guarantee weight information;

[0085] Step S43, the credit score, risk level, credit limit weight information, interest rate weight information and guarantee weight information are jointly generated as credit reference information and output, thereby completing the assessment of the credit limit risk of the enterprise to be assessed.

[0086] Furthermore, for different credit coefficients, different credit scores and risk levels are pre-set to correspond to them, and different credit scores and risk levels correspond to different credit limit information, interest rate information and guarantee information. That is, there is a first corresponding relationship between the credit coefficient and the credit score and risk level, a second corresponding relationship between the credit score and the credit limit information, interest rate information and guarantee information, and a third corresponding relationship between the risk level and the credit limit information, interest rate information and guarantee information. After the credit coefficient of the enterprise to be evaluated is analyzed by the industry evaluation module, the credit score and risk level corresponding to the credit coefficient are first found from the first corresponding relationship, and then the credit limit information, interest rate information and guarantee information corresponding to the credit score are found from the second corresponding relationship as the first credit limit information, the first interest rate information and the first guarantee information, and the credit limit information, interest rate information and guarantee information corresponding to the risk level are found from the third corresponding relationship as the second credit limit information, the second interest rate information and the second guarantee information.

[0087] Further, the credit weight and risk weight pre-set for the credit score and risk level are searched, and the first credit limit information and the second credit limit information are weighted according to the credit weight and the risk weight to obtain credit limit weight information, the first interest rate information and the second interest rate information are weighted according to the credit weight and the risk weight to obtain interest rate weight information, and the first guarantee information and the second guarantee information are weighted according to the credit weight and the risk weight to obtain guarantee weight information. Then, the data group of the credit score, risk level, credit limit weight information, interest rate weight information and guarantee weight information is generated as a credit reference information output, and the credit limit risk assessment of the enterprise to be assessed is completed.

[0088] The enterprise credit line risk assessment method implemented in this embodiment collects enterprise information corresponding to the enterprise to be assessed, and the collected enterprise information at least includes enterprise basic information, enterprise financial report, enterprise tax record, industry dynamic information, and financing investment information; and pre-processes the collected enterprise information to remove noise information and abnormal information therein to obtain the information to be assessed; and extracts the credit-related data in the information to be assessed according to the preset credit keywords; and then analyzes and evaluates the credit keywords and credit-related data according to the industry assessment module corresponding to the enterprise to be assessed in the preset neural network model, obtains the credit coefficient of the enterprise to be assessed, and evaluates the credit line risk of the enterprise to be assessed according to the credit coefficient. In this way, various types of information related to the enterprise are collected with the help of crawler technology, so that the collected information is more comprehensive, and then the credit-related data obtained after cleaning, screening and extraction of various types of information are analyzed and evaluated by the preset neural network model formed by training a large amount of sample data, so that the generated credit coefficient for assessing the enterprise credit line risk is more accurate, and a comprehensive and accurate assessment of the enterprise credit line risk is achieved.

[0089] For further information, please refer to Figure 2 Based on the first embodiment of the enterprise credit line risk assessment method of the present invention, a second embodiment of the enterprise credit line risk assessment method of the present invention is proposed.

[0090] The difference between the second embodiment of the enterprise credit line risk assessment method and the first embodiment of the enterprise credit line risk assessment method is that before the step of analyzing and assessing the credit keywords and credit-related data according to the industry assessment module corresponding to the enterprise to be assessed in the preset neural network model, the step includes:

[0091] Step S50, collecting credit training samples corresponding to the enterprise to be evaluated, and training the initial evaluation module corresponding to the enterprise to be evaluated in the preset initial network model according to the credit training samples;

[0092] Step S60, when the training duration reaches a preset duration, the loss function value corresponding to the initial evaluation module is calculated, and it is determined whether the calculated loss function value is less than a preset loss threshold;

[0093] Step S70, if it is less than the preset loss threshold, then the training of the initial assessment module is terminated, and the initial assessment module is generated as the industry assessment module;

[0094] Step S80, if the loss function value is not less than the preset loss threshold, executing the step of training the initial evaluation module corresponding to the enterprise to be evaluated in the preset initial network model according to the credit training sample until the loss function value is less than the preset loss threshold;

[0095] Step S90, when all modules included in the preset initial network model have completed training, the preset initial network model is generated as the preset neural network model.

[0096] Understandably, the preset neural network model contains multiple industry assessment modules corresponding to different industries, so the training of the preset neural network model also needs to be classified and trained based on the enterprise data of different industries. Among them, for the industry assessment model corresponding to the enterprise to be assessed, first collect the credit training samples corresponding to the preset assessment enterprise, and then train the initial assessment module corresponding to the enterprise to be assessed in the preset neural network model based on the credit training samples. The initial assessment module is an initial module pre-set for assessing the industry corresponding to the enterprise to be assessed. The duration of statistical training is calculated, and when the statistical training duration reaches the preset duration, the loss function value corresponding to the initial assessment module is calculated.

[0097] The specific calculation formula can be found in the following formula (2).

[0098]

[0099] Wherein, L represents the loss function value, n represents the number of credit training samples, yi represents the training value obtained after training of the i-th credit training sample, and yi0 represents the reference value corresponding to the i-th credit training sample.

[0100] Furthermore, in order to reflect the loss size of the loss function value, a preset loss threshold is pre-set, and the loss function value is compared with the preset loss threshold to determine whether the loss function value is less than the preset loss threshold. If it is less than the preset loss threshold, it means that the error between the training value obtained by the initial evaluation module for processing the credit training sample and the reference value corresponding to the credit training sample is small, and the initial evaluation module has good processing performance, so the training of the initial evaluation module is terminated and it is generated as an industry evaluation module in the preset neural network model for evaluating the industry in which the enterprise to be evaluated is located. On the contrary, if the loss function value is determined to be not less than the preset loss threshold through comparison, it means that the processing performance of the initial evaluation module is poor. At this time, the initial evaluation module is re-trained based on the credit training sample, and the new loss function value is recalculated until the calculated loss function value is less than the preset loss threshold, and the industry evaluation module in the preset neural network model is generated.

[0101] It should be noted that, because the preset neural network model includes modules of multiple industries, the modules of each industry can be trained according to the above method, and different loss function value calculation methods can be set for different modules, so that the loss function value is more matched with the module, ensuring the accuracy of each module. After each module included in the preset neural network model has completed its own training, the preset initial network model forms the preset neural network model.

[0102] Furthermore, in order to ensure the accuracy of each industry assessment module in the preset neural network model, the data analyzed and assessed by the industry assessment module can be used as training samples to update and train the preset neural network model. Specifically, the step of assessing the credit line risk of the enterprise to be assessed according to the credit coefficient includes:

[0103] Step a1, updating the credit training sample according to the credit keyword and the credit associated data, and generating an update rate of the credit training sample;

[0104] Step a2, determining whether the update rate is greater than a preset update threshold, if so, executing the step of training the initial assessment module corresponding to the enterprise to be assessed in the preset initial network model according to the credit training sample based on the updated credit training sample.

[0105] Furthermore, after completing the risk assessment of the credit limit of the enterprise to be assessed based on the credit coefficient, the credit keywords and their corresponding credit-related data are added to the credit training samples, and the credit training samples are updated, so as to update and train the initial assessment module corresponding to the enterprise to be assessed in the preset neural network model based on the updated credit training samples.

[0106] It should be noted that if the initial evaluation module is updated for each credit keyword and its corresponding credit-related data of each enterprise to be evaluated, the system overhead will be large and the update strength will be small. Therefore, after adding the credit keyword and its corresponding credit-related data to the credit training sample, the update amount of the credit training sample is counted, and the update amount is the number of new credit training samples added. Then, the update rate of the credit training sample is generated based on the update amount, that is, the ratio of the new credit training samples added to the original credit training samples.

[0107] Furthermore, in order to indicate the amount of updates, a preset update threshold is pre-set, and the generated update rate is compared with the preset update threshold to determine whether the update rate is greater than the preset update threshold. If it is greater, it means that there are more update samples in the credit training sample. At this time, the initial assessment module of the preset initial network model and the enterprise to be assessed is updated and trained based on the updated credit training sample to enrich the training samples of the initial assessment module and improve its accuracy. On the contrary, if it is determined through comparison that the update rate is less than the preset update threshold, the initial assessment module is not updated and trained until the update rate is greater than the preset update threshold through continued evaluation of the initial assessment module. In addition, this type of update training can also be set for other modules in the preset neural network model, which will not be elaborated here.

[0108] This embodiment sets the preset neural network model into multiple modules corresponding to different industries, and generates each industry assessment module through training with a large number of credit training samples in each industry, so that each industry assessment module can accurately assess the credit line risk of the enterprise corresponding to the respective industry. In addition, the assessment modules of each industry are updated and trained through the actual assessment data, further improving the accuracy of the credit line risk assessment of the enterprise.

[0109] In addition, the present invention provides a corporate credit risk assessment system, please refer to Figure 3 , Figure 3 This is a module diagram of the first embodiment of the enterprise credit line risk assessment system of the present invention.

[0110] Specifically, the enterprise credit line risk assessment system in this embodiment includes a data acquisition module 10, a preprocessing module 20, an extraction module 30 and an assessment module 40, wherein:

[0111] The collection module 10 is used to collect enterprise information corresponding to the enterprise to be evaluated, and the enterprise information includes but is not limited to basic enterprise information, enterprise financial reports, enterprise tax records, industry dynamic information, and financing and investment information;

[0112] The enterprise credit line risk assessment method of this embodiment is applied to the enterprise credit line risk assessment system. When an enterprise needs to initiate a financial loan from a bank or other financial institution, the bank or other financial institution initiates an assessment request to the enterprise credit line risk assessment system. After receiving the assessment request, the enterprise credit line risk assessment system reads the enterprise ID (Identification) carried therein, determines the specific enterprise that needs to undergo credit information assessment based on the enterprise ID, and uses the enterprise that needs to undergo credit information assessment as the enterprise to be assessed.

[0113] Furthermore, when the enterprise credit line risk assessment system initiates the enterprise credit line risk assessment for the enterprise to be assessed, its collection module first collects various enterprise information related to the enterprise to be assessed from various official websites, news websites, social platforms and other channels by calling API, web crawlers and other technologies. Among them, according to the nature of the website from which the enterprise information comes and the category of the information itself, the collected enterprise information at least includes basic enterprise information indicating the nature and scale of the enterprise, enterprise financial reports, enterprise tax information, industry dynamic information and financing investment information.

[0114] A preprocessing module 20 is used to preprocess the enterprise information, remove noise information and abnormal information in the enterprise information, and obtain information to be evaluated;

[0115] Understandably, the accuracy of various enterprise information collected varies, so the collected enterprise information needs to be preprocessed through the preprocessing module. Preprocessing includes data cleaning, removing repeated noise information that is irrelevant to the enterprise to be evaluated, and screening out abnormal information. Among them, the screening of abnormal information can be based on the consistency between the various information in the collected enterprise information. The consistency can be represented by the contradiction between the same type of data from different websites. If there is a contradiction between the same type of data, it means that the consistency between the data is low, and accurate data needs to be screened out. If there is no contradiction between the various types of data, it means that the consistency between the data is high and the data is relatively accurate. The order of screening is official data first, that is, official data shall prevail. After eliminating data irrelevant to the enterprise, removing duplicates, and screening out accurate data, the information to be evaluated related to the enterprise to be evaluated is obtained.

[0116] The extraction module 30 is used to extract the credit-related data in the information to be evaluated according to the preset credit keywords.

[0117] Further, banks and corporate financial institutions can set keywords related to the risk of credit lines that they are concerned about, based on their past experience in risk assessment of corporate credit lines, as the preset credit keywords. The extraction module in the corporate credit line risk assessment system performs text recognition on the information to be evaluated according to the preset credit keywords, and extracts the text related to the corporate credit line in the information to be evaluated through text recognition as the credit-related data corresponding to the enterprise to be evaluated. Specifically, the extraction module includes:

[0118] A determination unit 31, configured to determine the number of single Chinese characters included in the preset credit keywords, and use each character in the data to be evaluated as the first character, and perform word segmentation processing according to the number to obtain multiple keywords;

[0119] A comparison unit 32, configured to calculate the similarity values between the preset credit keywords and the multiple keywords respectively, and compare the similarity values to determine the maximum value among the similarity values;

[0120] A search unit 33, configured to search for the target keyword corresponding to the maximum value among the keywords, and search for the value corresponding to the target keyword in the information to be evaluated as the credit-related data.

[0121] Furthermore, the determination unit in the extraction module counts the number of single Chinese characters included in the preset credit keywords, and performs word segmentation processing on the data to be evaluated according to this number. After removing the numbers from the data to be evaluated, each character in it is used as the first character and segmented according to the number to obtain multiple keywords. For example, if the counted number is 3 and the data to be evaluated is "In the first half of 2024, the company achieved a total income of RMB A", then "上", "半", "年", "公", "司", "实", "现", "了", "总", "收", "入", "人", "民", "币" are used as the first characters respectively, and segmented according to the number 3 to obtain multiple keywords such as "上半年", "半年公", "年公司"... Among them, in order to perform more accurate word segmentation, the modal particles and conjunctions in the data to be evaluated can be removed before word segmentation. For example, the "了" in the above data to be evaluated. After removal, word segmentation is performed with each character as the first character to obtain multiple more accurate keywords.

[0122] Further, a vector dictionary and a calculation formula for calculating the word segmentation similarity are preset in the corporate credit line risk assessment system. The word vectors corresponding to the preset credit keywords and each keyword are found from the vector dictionary, and the comparison unit calculates the similarity values between the word vectors of the preset credit keywords and the word vectors of each keyword respectively through the calculation formula. The specific calculation formula can refer to the above formula (1) and will not be elaborated here.

[0123] Furthermore, the comparison unit compares each similarity value to determine the maximum value. Then, the search unit searches for the keyword that generates the maximum value among the keywords as the target keyword, and searches for the numerical value corresponding to the target keyword in the information to be evaluated as the credit-related data. For example, for the above-mentioned information to be evaluated "The company achieved a total revenue of RMB A in the first half of 2024", if the target keyword is "total revenue", then the numerical value A is the credit-related data. Among them, there can be multiple preset credit keywords set. After each information to be evaluated is extracted by each preset credit keyword, multiple credit-related data related to the enterprise to be evaluated are obtained.

[0124] The evaluation module 40 is used to analyze and evaluate the credit keywords and credit-related data according to the industry evaluation module corresponding to the enterprise to be evaluated in the preset neural network model, obtain the credit coefficient of the enterprise to be evaluated, and evaluate the credit line risk of the enterprise to be evaluated based on the credit coefficient.

[0125] Furthermore, a preset neural network model is pre-formed through training in the enterprise credit line risk assessment system. And, in order to better reflect the characteristics of enterprises in different industries, the preset neural network model can be trained according to the industry during training, so that the preset neural network model contains multiple industry assessment modules, and each industry assessment module corresponds to an industry. For the enterprise to be assessed, first determine the industry in which it is located, and then find the industry assessment module corresponding to the industry in the preset neural network model as the industry assessment module corresponding to the enterprise to be assessed. The assessment module analyzes and evaluates the credit keywords and the credit-related data corresponding to each credit keyword based on the industry assessment module to obtain the credit coefficient of the enterprise to be assessed. The credit coefficient represents the degree of loan risk of the enterprise to be assessed, and can be used to evaluate the credit line risk of the enterprise to be assessed.

[0126] Wherein, the evaluation module also includes:

[0127] An acquisition unit 41 is used to search for a credit score and a risk level corresponding to the credit coefficient, and acquire first credit limit information, first interest rate information, and first guarantee information corresponding to the credit score, and second credit limit information, second interest rate information, and second guarantee information corresponding to the risk level;

[0128] The weighting unit 42 is used to find the credit weight and risk weight corresponding to the credit score and risk level, and perform weighted processing on the first credit limit information, the first interest rate information, the first guarantee information, the second credit limit information, the second interest rate information, and the second guarantee information according to the credit weight and risk weight to obtain credit limit weight information, interest rate weight information, and guarantee weight information;

[0129] The generating unit 43 is used to generate the credit score, risk level, credit limit weight information, interest rate weight information and guarantee weight information into credit reference information and output it, so as to complete the credit limit risk assessment of the enterprise to be assessed.

[0130] Furthermore, for different credit coefficients, different credit scores and risk levels are pre-set to correspond to them, and different credit scores and risk levels correspond to different credit limit information, interest rate information and guarantee information. That is, there is a first corresponding relationship between the credit coefficient and the credit score and risk level, a second corresponding relationship between the credit score and the credit limit information, interest rate information and guarantee information, and a third corresponding relationship between the risk level and the credit limit information, interest rate information and guarantee information. After the industry evaluation module analyzes the credit coefficient of the enterprise to be evaluated, the acquisition unit first finds the credit score and risk level corresponding to the credit coefficient from the first corresponding relationship, and then finds the credit limit information, interest rate information and guarantee information corresponding to the credit score from the second corresponding relationship as the first credit limit information, the first interest rate information and the first guarantee information, and finds the credit limit information, interest rate information and guarantee information corresponding to the risk level from the third corresponding relationship as the second credit limit information, the second interest rate information and the second guarantee information.

[0131] Further, the weighting unit searches for the credit weight and risk weight pre-set for the credit score and risk level, and performs weighted processing on the first credit limit information and the second credit limit information according to the credit weight and the risk weight, respectively, to obtain credit limit weight information, performs weighted processing on the first interest rate information and the second interest rate information according to the credit weight and the risk weight, respectively, to obtain interest rate weight information, and performs weighted processing on the first guarantee information and the second guarantee information according to the credit weight and the risk weight, respectively, to obtain guarantee weight information. Then, the generating unit generates the data group of the credit score, risk level, credit limit weight information, interest rate weight information and guarantee weight information as credit reference information output, and completes the assessment of the credit limit risk of the enterprise to be assessed.

[0132] The enterprise credit line risk assessment system implemented in this embodiment collects enterprise information corresponding to the enterprise to be assessed through the collection module, and the collected enterprise information at least includes enterprise basic information, enterprise financial report, enterprise tax record, industry dynamic information, and financing investment information; and pre-processes the collected enterprise information through the pre-processing module to remove the noise information and abnormal information therein to obtain the information to be assessed; and the extraction module extracts the credit-related data in the information to be assessed according to the preset credit keywords; and then the assessment module analyzes and evaluates the credit keywords and credit-related data according to the industry assessment module corresponding to the enterprise to be assessed in the preset neural network model, obtains the credit coefficient of the enterprise to be assessed, and evaluates the credit line risk of the enterprise to be assessed according to the credit coefficient. In this way, various types of information related to the enterprise are collected with the help of crawler technology, so that the collected information is more comprehensive, and then the credit-related data obtained after cleaning, screening and extraction of various types of information are analyzed and evaluated by the preset neural network model formed by training a large amount of sample data, so that the generated credit coefficient for assessing the enterprise credit line risk is more accurate, and a comprehensive and accurate assessment of the enterprise credit line risk is achieved.

[0133] For further information, please refer to Figure 4 Based on the first embodiment of the enterprise credit line risk assessment system of the present invention, a second embodiment of the enterprise credit line risk assessment system of the present invention is proposed.

[0134] The difference between the second embodiment of the enterprise credit line risk assessment system and the first embodiment of the enterprise credit line risk assessment system is that the enterprise credit line risk assessment system further includes:

[0135] The training module 50 is used to collect the credit training samples corresponding to the enterprise to be evaluated, and train the initial evaluation module corresponding to the enterprise to be evaluated in the preset initial network model according to the credit training samples;

[0136] A calculation module 60 is used to calculate the loss function value corresponding to the initial evaluation module when the training duration reaches a preset duration, and determine whether the calculated loss function value is less than a preset loss threshold;

[0137] A generating module 70, when the loss function value is less than a preset loss threshold, ends the training of the initial evaluation module and generates the initial evaluation module as the industry evaluation module;

[0138] An execution module 80 is used for executing the step of training the initial evaluation module corresponding to the enterprise to be evaluated in the preset initial network model according to the credit training sample when the loss function value is not less than the preset loss threshold, until the loss function value is less than the preset loss threshold;

[0139] The generating module 70 is further configured to generate the preset initial network model into the preset neural network model when all modules included in the preset initial network model have completed training.

[0140] Understandably, the preset neural network model includes multiple industry assessment modules corresponding to different industries, so the training of the preset neural network model also needs to be classified and trained based on the enterprise data of different industries. Among them, for the industry assessment model corresponding to the enterprise to be assessed, the training module first collects the credit training samples corresponding to the preset assessment enterprise, and then trains the initial assessment module corresponding to the enterprise to be assessed in the preset neural network model based on the credit training samples. The initial assessment module is an initial module pre-set for assessing the industry corresponding to the enterprise to be assessed. The duration of statistical training, when the statistical training duration reaches the preset duration, the loss function value corresponding to the initial assessment module is calculated by the calculation module. The specific calculation formula can be found in the above formula (2), which will not be repeated here.

[0141] Furthermore, in order to reflect the loss size of the loss function value, a preset loss threshold is pre-set, and the loss function value is compared with the preset loss threshold to determine whether the loss function value is less than the preset loss threshold. If it is less than the preset loss threshold, it means that the error between the training value obtained by the initial evaluation module for processing the credit training sample and the reference value corresponding to the credit training sample is small, and the initial evaluation module has good processing performance, so the training of the initial evaluation module is terminated, and the generation module generates it as an industry evaluation module in the preset neural network model for evaluating the industry in which the enterprise to be evaluated is located. On the contrary, if the loss function value is determined to be not less than the preset loss threshold through comparison, it means that the processing performance of the initial evaluation module is poor. At this time, the execution module continues to re-train the initial evaluation module based on the credit training sample and recalculate the new loss function value until the calculated loss function value is less than the preset loss threshold, and the industry evaluation module in the preset neural network model is generated.

[0142] It should be noted that, because the preset neural network model includes modules of multiple industries, the modules of each industry can be trained according to the above method, and different loss function value calculation methods can be set for different modules, so that the loss function value is more matched with the module, ensuring the accuracy of each module. After each module included in the preset neural network model has completed its own training, the generation module will preset the initial network model to form a preset neural network model.

[0143] Furthermore, in order to ensure the accuracy of each industry assessment module in the preset neural network model, the data analyzed and evaluated by the industry assessment module can be used as training samples to update and train the preset neural network model. Specifically, the enterprise credit line risk assessment system also includes:

[0144] An updating module b1, used for updating the credit training sample according to the credit keyword and the credit associated data, and generating an update rate of the credit training sample;

[0145] The judgment module b2 is used to judge whether the update rate is greater than the preset update threshold. If it is greater than the preset update threshold, the step of training the initial evaluation module corresponding to the enterprise to be evaluated in the preset initial network model according to the credit training sample is executed based on the updated credit training sample.

[0146] Furthermore, after completing the risk assessment of the credit limit of the enterprise to be assessed based on the credit coefficient, the update module adds the credit keywords and their corresponding credit-related data to the credit training samples, updates the credit training samples, and updates the initial assessment module corresponding to the enterprise to be assessed in the preset neural network model based on the updated credit training samples.

[0147] It should be noted that if the initial evaluation module is updated for each credit keyword and its corresponding credit-related data of each enterprise to be evaluated, the system overhead will be large and the update strength will be small. Therefore, after adding the credit keyword and its corresponding credit-related data to the credit training sample, the update amount of the credit training sample is counted, and the update amount is the number of new credit training samples added. Then, the update rate of the credit training sample is generated based on the update amount, that is, the ratio of the new credit training samples added to the original credit training samples.

[0148] Furthermore, in order to indicate the amount of updates, a preset update threshold is pre-set, and the judgment module compares the generated update rate with the preset update threshold to determine whether the update rate is greater than the preset update threshold. If it is greater, it means that there are more update samples in the credit training sample. At this time, the initial assessment module of the preset initial network model and the enterprise to be assessed is updated and trained based on the updated credit training sample to enrich the training samples of the initial assessment module and improve its accuracy. On the contrary, if it is determined through comparison that the update rate is less than the preset update threshold, the initial assessment module will not be updated and trained until the update rate is greater than the preset update threshold through continued evaluation of the initial assessment module. In addition, this type of update training can also be set for other modules in the preset neural network model, which will not be elaborated here.

[0149] This embodiment sets the preset neural network model into multiple modules corresponding to different industries, and generates each industry assessment module through training with a large number of credit training samples in each industry, so that each industry assessment module can accurately assess the credit line risk of the enterprise corresponding to the respective industry. In addition, the assessment modules of each industry are updated and trained through the actual assessment data, further improving the accuracy of the credit line risk assessment of the enterprise.

[0150] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims. All equivalent structures or equivalent process changes made using the contents of the specification and drawings of the present invention, or directly or indirectly used in other related technical fields, are protected by the present invention.

Claims

1. A method for assessing the risk of an enterprise's credit line, characterized in that: The enterprise credit line risk assessment method includes: Collect enterprise information corresponding to the enterprise to be evaluated, including but not limited to basic enterprise information, enterprise financial reports, enterprise tax records, industry dynamics information, and financing and investment information; Preprocessing the enterprise information to remove noise information and abnormal information in the enterprise information to obtain information to be evaluated; Extracting credit-related data from the information to be evaluated according to preset credit keywords; According to the industry assessment module corresponding to the enterprise to be assessed in the preset neural network model, the credit keywords and credit-related data are analyzed and evaluated to obtain the credit coefficient of the enterprise to be assessed, and the credit line risk of the enterprise to be assessed is evaluated based on the credit coefficient.

2. The enterprise credit line risk assessment method according to claim 1, characterized in that: The step of evaluating the credit line risk of the enterprise to be evaluated according to the credit coefficient comprises: Finding a credit score and a risk level corresponding to the credit coefficient, and obtaining first credit limit information, first interest rate information, and first guarantee information corresponding to the credit score, and second credit limit information, second interest rate information, and second guarantee information corresponding to the risk level; Finding the credit weight and risk weight corresponding to the credit score and risk level, and weighting the first credit limit information, the first interest rate information, the first guarantee information, the second credit limit information, the second interest rate information, and the second guarantee information according to the credit weight and risk weight to obtain credit limit weight information, interest rate weight information, and guarantee weight information; The credit score, risk level, credit line weight information, interest rate weight information and guarantee weight information are jointly generated as credit reference information and output, thereby completing the assessment of the credit line risk of the enterprise to be assessed.

3. The enterprise credit line risk assessment method according to claim 1, characterized in that: The step of extracting the credit-related data in the information to be evaluated according to the preset credit keywords includes: Determine the number of single Chinese characters contained in the preset credit keyword, and use each character in the data to be evaluated as the first character, perform word segmentation processing according to the number, and obtain multiple keywords; Calculate the similarity values ​​between the preset credit keyword and the multiple keywords respectively, and compare the similarity values ​​to determine the maximum value among the similarity values, wherein the similarity value calculation formula is: Where S represents the similarity value, W represents the word vector of the preset credit keyword, and Wi represents the word vector of the i-th keyword; A target keyword corresponding to the maximum value is searched in each of the keywords, and a value corresponding to the target keyword is searched in the information to be evaluated as the credit-related data.

4. The enterprise credit line risk assessment method according to any one of claims 1 to 3, characterized in that: The step of analyzing and evaluating the credit keywords and credit-related data according to the industry evaluation module corresponding to the enterprise to be evaluated in the preset neural network model includes: Collecting credit training samples corresponding to the enterprise to be evaluated, and training the initial evaluation module corresponding to the enterprise to be evaluated in the preset initial network model according to the credit training samples; When the training duration reaches the preset duration, the loss function value corresponding to the initial evaluation module is calculated, and it is determined whether the calculated loss function value is less than the preset loss threshold, wherein the calculation formula of the loss function value is: Wherein, L represents the loss function value, n represents the number of credit training samples, yi represents the training value obtained by training the i-th credit training sample, and yi0 represents the reference value corresponding to the i-th credit training sample; If it is less than the preset loss threshold, the training of the initial assessment module is terminated, and the initial assessment module is generated as the industry assessment module; If the loss function value is not less than the preset loss threshold, executing the step of training the initial evaluation module corresponding to the enterprise to be evaluated in the preset initial network model according to the credit training sample until the loss function value is less than the preset loss threshold; When all modules included in the preset initial network model have completed training, the preset initial network model is generated as the preset neural network model.

5. The enterprise credit line risk assessment method according to claim 4, characterized in that: The step of evaluating the credit line risk of the enterprise to be evaluated according to the credit coefficient includes: According to the credit keyword and the credit associated data, the credit training sample is updated, and an update rate of the credit training sample is generated; Determine whether the update rate is greater than a preset update threshold. If so, perform a step of training an initial assessment module corresponding to the enterprise to be assessed in a preset initial network model based on the updated credit training sample.

6. An enterprise credit line risk assessment system, characterized in that: The enterprise credit line risk assessment system includes: A collection module, used to collect enterprise information corresponding to the enterprise to be evaluated, including but not limited to basic enterprise information, enterprise financial reports, enterprise tax records, industry dynamic information, and financing and investment information; A preprocessing module, used to preprocess the enterprise information, remove noise information and abnormal information in the enterprise information, and obtain information to be evaluated; An extraction module, used to extract the credit-related data in the information to be evaluated according to preset credit keywords; An evaluation module is used to analyze and evaluate the credit keywords and credit-related data according to the industry evaluation module corresponding to the enterprise to be evaluated in the preset neural network model, obtain the credit coefficient of the enterprise to be evaluated, and evaluate the credit line risk of the enterprise to be evaluated based on the credit coefficient.

7. The enterprise credit line risk assessment system according to claim 6, characterized in that: The evaluation module also includes: an acquisition unit, configured to search for a credit score and a risk level corresponding to the credit coefficient, and acquire first credit limit information, first interest rate information, and first guarantee information corresponding to the credit score, and second credit limit information, second interest rate information, and second guarantee information corresponding to the risk level; a weighting unit, configured to search for a credit weight and a risk weight corresponding to the credit score and the risk level, and perform weighted processing on the first credit limit information, the first interest rate information, the first guarantee information, the second credit limit information, the second interest rate information, and the second guarantee information according to the credit weight and the risk weight, so as to obtain credit limit weight information, interest rate weight information, and guarantee weight information; The generating unit is used to generate the credit score, risk level, credit limit weight information, interest rate weight information and guarantee weight information into credit reference information and output it, so as to complete the credit limit risk assessment of the enterprise to be assessed.

8. The enterprise credit line risk assessment system according to claim 6, characterized in that: The extraction module comprises: A determination unit, used to determine the number of single Chinese characters contained in the preset credit keyword, and use each character in the data to be evaluated as the first character, perform word segmentation processing according to the number, and obtain multiple keywords; The comparison unit is used to calculate the similarity values ​​between the preset credit keyword and the plurality of keywords respectively, and compare the similarity values ​​to determine the maximum value among the similarity values, wherein the similarity value calculation formula is: Where S represents the similarity value, W represents the word vector of the preset credit keyword, and Wi represents the word vector of the i-th keyword; A searching unit is used to search for a target keyword corresponding to the maximum value in each of the keywords, and to search for a value corresponding to the target keyword in the information to be evaluated as the credit-related data.

9. The enterprise credit line risk assessment system according to any one of claims 6 to 8, characterized in that: The enterprise credit line risk assessment system also includes: A training module, used for collecting credit training samples corresponding to the enterprise to be evaluated, and training the initial evaluation module corresponding to the enterprise to be evaluated in the preset initial network model according to the credit training samples; The calculation module is used to calculate the loss function value corresponding to the initial evaluation module when the training duration reaches the preset duration, and determine whether the calculated loss function value is less than the preset loss threshold, wherein the calculation formula of the loss function value is: Wherein, L represents the loss function value, n represents the number of credit training samples, yi represents the training value obtained by training the i-th credit training sample, and yi0 represents the reference value corresponding to the i-th credit training sample; A generation module, when the loss function value is less than a preset loss threshold, ends the training of the initial evaluation module and generates the initial evaluation module as the industry evaluation module; An execution module, configured to execute, when the loss function value is not less than a preset loss threshold, a step of training an initial evaluation module corresponding to the enterprise to be evaluated in a preset initial network model according to the credit training sample, until the loss function value is less than the preset loss threshold; The generating module is further used to generate the preset initial network model into the preset neural network model when all modules included in the preset initial network model have completed training.

10. The enterprise credit line risk assessment system according to claim 9, characterized in that: The enterprise credit line risk assessment system also includes: An updating module, used for updating the credit training sample according to the credit keyword and the credit associated data, and generating an update rate of the credit training sample; The judgment module is used to judge whether the update rate is greater than a preset update threshold. If it is greater than the preset update threshold, the step of training the initial evaluation module corresponding to the enterprise to be evaluated in the preset initial network model according to the credit training sample is executed based on the updated credit training sample.

Images