Method and device for selecting target enterprise based on classification model

Abstract

The invention provides a method and a device for selecting a target enterprise based on a classification model. The method comprises the following steps: accessing a target data platform in advance to obtain credit information of multiple benchmarking enterprises meeting credit acquisition conditions and first candidate enterprises not meeting preset credit investigation conditions; the first candidate enterprise waits to enter a credit auditing node; sampling and selecting a first part of benchmarking enterprises and the first candidate enterprises from the benchmarking enterprises to form a sample candidate enterprise set, forming a sample benchmarking enterprise set by the remaining second part of benchmarking enterprises, and training an admission classification model to obtain confidence coefficients of the first part of benchmarking enterprises and the first candidate enterprises belonging to admission categories; and according to the confidence coefficients of the first part of benchmarking enterprises and the first candidate enterprises, selecting a target enterprise to enter a credit auditing node from the first candidate enterprises, thereby automatically selecting high-quality small and micro enterprises to enter a subsequent credit auditing node, and objectively and efficiently solving the cold start problem of enterprise credit investigation white user admission.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method and device for selecting a target enterprise based on a classification model. Background Art

[0002] Small and micro enterprises and self-employed businesses face difficulties and high costs in financing, partly because of the lack of financial data of small and micro enterprises and information asymmetry between banks and enterprises. Currently, for the access of small and micro enterprises, their corresponding financial indicators, tax payment indicators and other indicators must meet certain conditions before they can enter the pre-loan review stage, or directly lend to enterprises with good credit in the past.

[0003] These small and micro enterprises that are allowed to enter the admission list are often those with very good reputations or those that have already obtained loans. As for those with no credit record, due to insufficient credit-related data, they will basically be considered to not meet the preset credit conditions, and it will be difficult for them to be included in the admission list, and they will not be able to enjoy the support brought by inclusive finance.

[0004] In the existing technology, the entry assessment for these enterprises that do not meet the preset credit conditions often requires manual analysis and evaluation, which has strong subjective factors and low efficiency. Summary of the invention

[0005] In view of this, the purpose of this application is to provide a method, device, electronic device and medium for selecting target enterprises based on a classification model, which can automatically select high-quality small and micro enterprises to enter subsequent credit review nodes, with objective evaluation results and high evaluation efficiency, reducing repeated credit to enterprises that have already borrowed money and increasing the degree of automation of the credit review process.

[0006] The method for selecting a target enterprise based on a classification model provided in an embodiment of the present application is applied to a server or a terminal device, and the method includes:

[0007] Accessing the target data platform in advance, and obtaining credit information of multiple benchmarking enterprises that meet the credit acquisition conditions from the target data platform, wherein the credit information of the benchmarking enterprises includes government affairs data of the target type;

[0008] Acquiring from the target data platform the credit information of a first candidate enterprise whose credit information does not meet the preset credit conditions; the credit information of the first candidate enterprise also includes government affairs data of the target type; the first candidate enterprise is an enterprise to enter a credit review node;

[0009] A first part of benchmarking enterprises are sampled and selected from the benchmarking enterprises, the first part of benchmarking enterprises and the first candidate enterprise constitute a sample candidate enterprise set, and the remaining second part of benchmarking enterprises constitute a sample benchmarking enterprise set; the sample benchmarking enterprise set and the sample candidate enterprise set include government affairs data of target types corresponding to the enterprises; the category of the sample candidate enterprise set is the access category;

[0010] The access classification model constructed by training based on the sample benchmark enterprise set and the sample candidate enterprise set is obtained until the access classification model meets a preset training stop condition; the trained access classification model outputs the confidence that the first part of the benchmark enterprises and the first candidate enterprise belong to the access category;

[0011] According to the confidence of the benchmark enterprises in the first part and the confidence of the first candidate enterprise, a target enterprise to enter the credit review node is selected from the first candidate enterprises.

[0012] In some embodiments, the method for selecting a target enterprise based on a classification model further includes:

[0013] Processing the target type of government data to obtain multiple indicators of each type of government data in multiple time dimensions;

[0014] The target type of government data corresponding to the enterprises in the sample benchmark enterprise set and the sample candidate enterprise set are multiple indicators of each type of government data in multiple time dimensions.

[0015] In some embodiments, in the method for selecting a target enterprise based on a classification model, the target type of government data is processed to obtain multiple indicators of each type of government data in multiple time dimensions, including:

[0016] For the target type of government affairs data, calculating the mean of each type of government affairs data in different time dimensions;

[0017] Based on the mean of each type of government data in different time dimensions, the year-on-year growth rate of the mean of this type of government data in this time dimension is calculated.

[0018] In some embodiments, in the method for selecting a target enterprise based on a classification model, before training the constructed admission classification model based on the sample benchmark enterprise set and the sample candidate enterprise set, the method further includes:

[0019] Multiple indicators of each type of government data in multiple time dimensions are used as input variables to construct an access classification model based on a binary classification algorithm.

[0020] In some embodiments, in the method for selecting a target enterprise based on a classification model, the credit acquisition conditions include:

[0021] The type of loan obtained by the benchmark enterprise is a specified type, and the time when the loan is obtained by the benchmark enterprise is within a first preset period, and no specified risk event occurs to the benchmark enterprise within a second preset period.

[0022] In some embodiments, in the method for selecting a target enterprise based on a classification model, selecting a target enterprise to enter a credit review node from the first candidate enterprises according to the confidence of the first part of benchmark enterprises and the confidence of the first candidate enterprise includes:

[0023] Determine the admission threshold of the first candidate enterprise based on the confidence of the first part of benchmarking enterprises;

[0024] Based on the admission threshold, selecting first candidate enterprises to be admitted whose confidence exceeds the admission threshold from the first candidate enterprises at one time;

[0025] A preset number of first candidate enterprises with higher confidence rankings are selected from the first candidate enterprises to be admitted as target enterprises entering the credit review node.

[0026] In some embodiments, in the method for selecting a target enterprise based on a classification model, after selecting a target enterprise to enter a credit review node from the first candidate enterprises, the method further includes:

[0027] When the target data platform detects a second candidate enterprise that does not meet the preset credit conditions, the credit information of the second candidate enterprise is obtained from the target data platform; the credit information of the second candidate enterprise also includes the government affairs data of the target type;

[0028] Processing the target type of government data in the credit information of the second candidate enterprise, and determining multiple indicators of multiple types of government data of the second candidate enterprise in multiple time dimensions;

[0029] Processing multiple indicators of multiple types of government affairs data of the second candidate enterprise in multiple time dimensions through the trained admission classification model, and outputting the confidence that the second candidate enterprise belongs to the admission category;

[0030] Based on the confidence of the second candidate enterprise, determine whether the second candidate enterprise enters the credit review node.

[0031] In some embodiments, a device for selecting a target enterprise based on a classification model is also provided, which is applied to a server or a terminal device, and includes:

[0032] A first acquisition module is used to pre-access the target data platform and acquire the credit information of a plurality of benchmark enterprises that meet the credit acquisition conditions from the target data platform, wherein the credit information of the benchmark enterprises includes government affairs data of the target type;

[0033] A second acquisition module is used to acquire, from the target data platform, the credit information of a first candidate enterprise whose credit information does not meet the preset credit conditions; the credit information of the first candidate enterprise also includes government affairs data of the target type; the first candidate enterprise is an enterprise to enter a credit review node;

[0034] A sampling module is used to sample and select a first part of benchmarking enterprises from the benchmarking enterprises, wherein the first part of benchmarking enterprises and the first candidate enterprise constitute a sample candidate enterprise set, and the remaining second part of benchmarking enterprises constitute a sample benchmarking enterprise set; the sample benchmarking enterprise set and the sample candidate enterprise set include government affairs data of target types corresponding to the enterprises; the category of the sample candidate enterprise set is an access category;

[0035] A training module is used to train the constructed admission classification model based on the sample benchmark enterprise set and the sample candidate enterprise set until the admission classification model meets a preset training stop condition, thereby obtaining a trained admission classification model; the trained admission classification model outputs the confidence that the first part of the benchmark enterprises and the first candidate enterprise belong to the admission category;

[0036] A selection module is used to select a target enterprise to enter the credit review node from the first candidate enterprises based on the confidence of the first part of benchmark enterprises and the confidence of the first candidate enterprises.

[0037] In some embodiments, an electronic device is also provided, including: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate through the bus, and when the machine-readable instructions are executed by the processor, the steps of the method for selecting a target enterprise based on a classification model are performed.

[0038] In some embodiments, a computer-readable storage medium is also provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for selecting a target enterprise based on a classification model are executed.

[0039] In an embodiment of the present application, a method and apparatus for selecting a target enterprise based on a classification model are specifically provided, which are applied to a server or a terminal device, and the method comprises: pre-accessing a target data platform, and obtaining credit information of a plurality of benchmarking enterprises that meet credit acquisition conditions from the target data platform, wherein the credit information of the benchmarking enterprises includes government affairs data of the target type; obtaining credit information of a first candidate enterprise whose credit information does not meet the preset credit conditions from the target data platform; the credit information of the first candidate enterprise also includes government affairs data of the target type; the first candidate enterprise is an enterprise to enter a credit review node; sampling and selecting a first part of benchmarking enterprises from the benchmarking enterprises, wherein the first part of benchmarking enterprises and the first candidate enterprise constitute a sample candidate enterprise set, and the remaining second part of benchmarking enterprises constitute a sample benchmarking enterprise set; the sample benchmarking enterprise set and the sample candidate enterprise set include government affairs data of the target type corresponding to the enterprise; the category of the sample candidate enterprise set is an access category; training a constructed access classification model based on the sample benchmarking enterprise set and the sample candidate enterprise set until the access classification is completed. The class model meets the preset training stop condition to obtain a trained access classification model; the trained access classification model outputs the confidence that the first part of the benchmark enterprises and the first candidate enterprises belong to the access category; according to the confidence of the first part of the benchmark enterprises and the confidence of the first candidate enterprises, the target enterprise to enter the credit review node is selected from the first candidate enterprises; firstly, the method can automatically select high-quality small and micro enterprises to enter the subsequent credit review node, the evaluation result is objective and the evaluation efficiency is high, the repeated credit to the enterprises that have already borrowed money is reduced, and the degree of automation of the credit review process is increased; secondly, by mixing the benchmark enterprises that do not meet the credit acquisition conditions into the first candidate enterprises that do not meet the preset credit conditions, the difference between the positive and negative sample sets of the classification model is reduced, and the generalization ability of the access classification model is improved. The access threshold of the first candidate enterprise is determined based on the confidence result of the benchmark enterprise, which improves the accuracy of the selection, thereby further expanding the scope of the target enterprises that can enter the credit review node, reducing the repeated credit to the enterprises that have already borrowed money, and solving the cold start problem of the access of credit-free enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 A flow chart showing a method for selecting a target enterprise based on a classification model according to an embodiment of the present application is shown;

[0042] Figure 2 A flow chart of a method for processing government affairs data of the target type according to an embodiment of the present application is shown;

[0043] Figure 3 A flow chart of a method for selecting a target enterprise to enter a credit review node from the first candidate enterprises according to an embodiment of the present application is shown;

[0044] Figure 4 A flowchart of another method for selecting a target enterprise based on a classification model according to an embodiment of the present application is shown;

[0045] Figure 5 A schematic diagram of the structure of the device for selecting a target enterprise based on a classification model according to an embodiment of the present application is shown;

[0046] Figure 6 A schematic structural diagram of an electronic device described in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0047] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of explanation and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn in real proportion. The flowchart used in this application shows the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowchart can be implemented out of sequence, and the steps without logical context can be reversed in order or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flowchart under the guidance of the content of the present application, or remove one or more operations from the flowchart.

[0048] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0049] It should be noted that the term "comprising" will be used in the embodiments of the present application to indicate the existence of the features declared thereafter, but does not exclude the addition of other features.

[0050] Small and micro enterprises and self-employed businesses face difficulties and high costs in financing, partly because of the lack of financial data of small and micro enterprises and information asymmetry between banks and enterprises. Currently, for the access of small and micro enterprises, their corresponding financial indicators, tax payment indicators and other indicators must meet certain conditions before they can enter the pre-loan review stage, or directly lend to enterprises with good credit in the past.

[0051] These small and micro enterprises that are allowed to enter the admission list are often those with very good reputations or those that have already obtained loans. As for those with no credit record, due to insufficient credit-related data, they will basically be considered to not meet the preset credit conditions, and it will be difficult for them to be included in the admission list, and they will not be able to enjoy the support brought by inclusive finance.

[0052] In the existing technology, the entry assessment for these enterprises that do not meet the preset credit conditions often requires manual analysis and evaluation, which has strong subjective factors and low efficiency.

[0053] Based on this, in an embodiment of the present application, a method and device for selecting target enterprises based on a classification model are specifically provided, which are applied to a server or a terminal device, and the method includes: pre-accessing a target data platform, and obtaining credit information of multiple benchmarking enterprises that meet credit acquisition conditions from the target data platform, wherein the credit information of the benchmarking enterprises includes government affairs data of the target type; obtaining credit information of a first candidate enterprise whose credit information does not meet the preset credit conditions from the target data platform; the credit information of the first candidate enterprise also includes government affairs data of the target type; the first candidate enterprise is an enterprise to enter a credit review node; a first part of benchmarking enterprises are sampled and selected from the benchmarking enterprises, the first part of benchmarking enterprises and the first candidate enterprises constitute a sample candidate enterprise set, and the remaining second part of benchmarking enterprises constitute a sample benchmarking enterprise set; the sample benchmarking enterprise set and the sample candidate enterprise set include government affairs data of the target type corresponding to the enterprise; the category of the sample candidate enterprise set is an access category; training a constructed access classification model based on the sample benchmarking enterprise set and the sample candidate enterprise set until the access classification model is constructed. The training classification model meets the preset training stop condition to obtain a trained access classification model; the trained access classification model outputs the confidence that the first part of the benchmark enterprises and the first candidate enterprises belong to the access category; according to the confidence of the first part of the benchmark enterprises and the confidence of the first candidate enterprises, the target enterprise to enter the credit review node is selected from the first candidate enterprises; firstly, the method can automatically select high-quality small and micro enterprises to enter the subsequent credit review node, the evaluation result is objective and the evaluation efficiency is high, the repeated credit to the enterprises that have been loaned is reduced, and the degree of automation of the credit review process is increased; secondly, by mixing the benchmark enterprises that do not meet the credit acquisition conditions into the first candidate enterprises that do not meet the preset credit conditions, the difference between the positive and negative sample sets of the classification model is reduced, the generalization ability of the access classification model is improved, and the access threshold of the first candidate enterprise is determined based on the confidence result of the benchmark enterprise, which improves the accuracy of the selection, thereby further expanding the scope of the target enterprises that can enter the credit review node, reducing the repeated credit to the enterprises that have been loaned, and solving the cold start problem of the access of the credit white enterprises.

[0054] Please refer to Figure 1 , Figure 1 A flow chart of the method for selecting a target enterprise based on a classification model according to an embodiment of the present application is shown; Figure 1 As shown, the method for selecting a target enterprise based on a classification model is applied to a server or a terminal device, and the method includes the following steps S101-S105:

[0055] S101, pre-accessing a target data platform, and obtaining credit information of a plurality of benchmarking enterprises that meet credit acquisition conditions from the target data platform, wherein the credit information of the benchmarking enterprises includes government affairs data of the target type;

[0056] S102, obtaining, from the target data platform, the credit information of a first candidate enterprise whose credit information does not meet the preset credit conditions; the credit information of the first candidate enterprise also includes government affairs data of the target type; the first candidate enterprise is an enterprise to enter a credit review node;

[0057] S103, sampling and selecting a first part of benchmarking enterprises from the benchmarking enterprises, wherein the first part of benchmarking enterprises and the first candidate enterprise constitute a sample candidate enterprise set, and the remaining second part of benchmarking enterprises constitute a sample benchmarking enterprise set; the sample benchmarking enterprise set and the sample candidate enterprise set include government affairs data of target types corresponding to the enterprises; the category of the sample candidate enterprise set is the access category;

[0058] S104, training the constructed admission classification model based on the sample benchmark enterprise set and the sample candidate enterprise set until the admission classification model meets a preset training stop condition, thereby obtaining a trained admission classification model; the trained admission classification model outputs the confidence that the first part of the benchmark enterprises and the first candidate enterprise belong to the admission category;

[0059] S105. Select a target enterprise to enter a credit review node from among the first candidate enterprises based on the confidence of the first part of benchmark enterprises and the confidence of the first candidate enterprise.

[0060] In the embodiments of this application, the enterprises are all specific types of enterprises, specifically small and micro enterprises; small and micro enterprises refer to those enterprises that are small in scale, flexible in operation, and usually in the start-up or growth stage. The specific standards of small and micro enterprises vary from industry to industry. Generally speaking, they need to meet one or more of the following conditions:

[0061] The number of employees is small. For example, some industries may stipulate that the number of employees is 300 or less.

[0062] The operating income is low. For example, some industries may stipulate that the annual operating income must be below a certain amount.

[0063] The total assets are relatively small. For example, some industries may stipulate that the total assets must be below a certain amount.

[0064] In other words, the specific type of enterprise is an enterprise whose enterprise information meets the preset enterprise type conditions.

[0065] In step S101, a target data platform is accessed in advance, and credit information of a plurality of benchmark enterprises that meet credit acquisition conditions is obtained from the target data platform, wherein the credit information of the benchmark enterprises includes government affairs data of the target type.

[0066] Here, the benchmark enterprise refers to the enterprise to which loans have been extended.

[0067] The data platform is a data platform of a credit review agency, such as a bank's risk control system, etc.

[0068] The benchmark enterprises for credit acquisition conditions are small and micro enterprises that have already been loaned and have good credit.

[0069] Specifically, in an embodiment of the present application, the credit acquisition conditions include: the type of loan obtained by the benchmark enterprise is a specified type, and the time when the loan is obtained by the benchmark enterprise is within a first preset period, and the benchmark enterprise has not experienced a specified risk event within a second preset period.

[0070] The type of loan obtained by the benchmark enterprise is a specified type. Specifically, the type of loan obtained by the benchmark enterprise is an inclusive small and micro enterprise loan.

[0071] The time when the benchmark enterprise obtained the loan is within the first preset period. For example, it needs to have obtained an inclusive small and micro enterprise loan in the past two years. Of course, the first preset period can also be in the past three years, in the past year, etc.

[0072] The benchmark enterprises have not experienced any designated risk events within the second preset period. Specifically, the enterprises that have obtained inclusive small and micro enterprise loans have not experienced any concerns, substandard, suspicious, losses, etc. after a period of performance.

[0073] Specifically, "Concern" means that the principal and interest of the loan are overdue (including after extension) for less than 90 days or interest is in arrears; "Substandard" means that the borrower's repayment ability has obvious problems and it is impossible to fully repay the principal and interest of the loan by relying solely on its normal operating income. Even if the guarantee is executed, it may cause certain losses; "Suspicious" means that the borrower cannot fully repay the principal and interest of the loan. Even if the guarantee is executed, it will definitely cause greater losses; "Loss" means that after taking all possible measures or all necessary legal procedures, the principal and interest still cannot be recovered, or only a very small part can be recovered.

[0074] Specifically, if the designated risk event occurs in the benchmark enterprise, a record will be made; if the designated risk event does not occur in the benchmark enterprise within the second preset period, that is, the designated risk event record information of the benchmark enterprise within the second preset period is not retrieved.

[0075] The type of loan obtained by the benchmark enterprise is the specified type, and the time when the benchmark enterprise obtained the loan is within the first preset period, indicating that the enterprise has successfully obtained an inclusive loan recently; and the benchmark enterprise has not experienced any specified risk events within the second preset period, further confirming that the company has performed well after obtaining the loan and has not experienced any situation that may increase the loan risk, indicating that its credit status is good. Therefore, these three conditions must be met at the same time to become a benchmark enterprise that meets the preset credit conditions and meets the credit acquisition conditions.

[0076] In step S102, the credit information of the first candidate enterprise whose credit information does not meet the preset credit conditions is obtained from the target data platform; the credit information of the first candidate enterprise also includes government affairs data of the target type; the first candidate enterprise is an enterprise to enter the credit review node;

[0077] If the credit information does not meet the preset credit conditions, the enterprise is considered to have no credit record.

[0078] In step S103, a first part of benchmarking enterprises are sampled and selected from the benchmarking enterprises. The first part of benchmarking enterprises and the first candidate enterprise constitute a sample candidate enterprise set, and the remaining second part of benchmarking enterprises constitute a sample benchmarking enterprise set; the sample benchmarking enterprise set and the sample candidate enterprise set include government affairs data of the target type corresponding to the enterprises; the category of the sample candidate enterprise set is the access category.

[0079] Specifically, the number of benchmark enterprises is set to P and the number of first candidate enterprises is set to N; S benchmark enterprises are sampled from the P benchmark enterprises, and the remaining PS benchmark enterprises with a label of 1 constitute a sample benchmark enterprise set (i.e., a seed sample enterprise set); the S benchmark enterprises and the N first candidate enterprises constitute a sample candidate enterprise set with a label of 0 (i.e., a non-seed sample enterprise set).

[0080] The label is 1, that is, the label is an access category.

[0081] The P benchmark enterprises are used as samples to train the admission classification model.

[0082] S benchmarking enterprises are sampled and selected from the P benchmarking enterprises. Specifically, S benchmarking enterprises are randomly sampled and selected from the P benchmarking enterprises.

[0083] S benchmarking enterprises are sampled from the P benchmarking enterprises, and the remaining PS benchmarking enterprises with label 1 constitute the seed sample enterprise set. The S benchmarking enterprises and N first candidate enterprises constitute the non-seed sample enterprise set with label 0. That is to say, the seed sample enterprise set is PS, and the label of the enterprises in the seed sample enterprise set is 1; the non-seed sample enterprise set is N+S, and the label of the enterprises in the non-seed sample enterprise set is 0.

[0084] The benchmark enterprises are taken as positive samples. However, in the access scenario in the embodiment of the present application, there are no enterprises as negative samples corresponding to the small and micro enterprises that have already been loaned and have good credit. There are only a large number of credit-free enterprises (first candidate enterprises). The purpose is to find enterprises with similar characteristics to the benchmark enterprises used as positive samples from a large number of credit-free enterprises.

[0085] In this way, if the traditional supervised machine learning model is used, it is impossible to establish an effective binary classifier due to the gap between positive and negative samples, and the unsupervised machine learning model will not produce satisfactory results.

[0086] Based on this, in the embodiment of the present application, by placing a portion of positive samples (benchmark enterprises) into the unlabeled sample set (first candidate enterprises), reliable credit-free enterprises with smaller differences from the positive samples can be more accurately identified, and credit-free enterprises with larger differences from the positive samples can be distinguished, thereby further expanding the scope of small and micro enterprises that can be approved and further improving the accuracy.

[0087] In other words, if there are no spy samples, and only P and N are used to build models, because the performance of the sample sets of P and N is very different, and their labels are companies that have received loans and companies that have not received loans, this binary classification model will be very effective in distinguishing between companies that have received loans and companies that have not received loans, but the generalization ability is very, very poor, and it is difficult to achieve the purpose of determining the allowed companies from the credit-free households as described in the embodiments of the present application.

[0088] In an embodiment of the present application, S benchmarking enterprises are sampled from the P benchmarking enterprises, and the value of S is determined according to a preset ratio; the preset ratio is between 10% and 20%. Exemplarily, 15% of the enterprise set S is sampled from P as the first part of benchmarking enterprises, and the remaining 85% are the second part of benchmarking enterprises.

[0089] In the embodiment of the present application, the sample benchmark enterprise set and the sample candidate enterprise set include government affairs data of the target type corresponding to the enterprise.

[0090] Specifically, the sample benchmark enterprise set and the sample candidate enterprise set are the government data indicators obtained by processing the target type of government data, specifically: multiple indicators of multiple types of government data in multiple time dimensions.

[0091] Specifically, the method for selecting a target enterprise based on a classification model described in the embodiment of the present application also includes:

[0092] Processing the target type of government data to obtain multiple indicators of each type of government data in multiple time dimensions;

[0093] The target type of government data corresponding to the enterprises in the sample benchmark enterprise set and the sample candidate enterprise set are multiple indicators of each type of government data in multiple time dimensions.

[0094] Exemplarily, the multiple types of government data include: value-added tax payment amount, income tax payment amount, employee basic pension insurance (unit payment) payment amount, provident fund contribution amount; the multiple time dimensions are the most recent three months, the most recent six months, and the most recent twelve months; the multiple indicators are the mean and the mean year-on-year growth rate.

[0095] Based on this, please refer to Figure 2 , processing the target type of government data to obtain multiple indicators of each type of government data in multiple time dimensions, including the following steps S201-S202:

[0096] S201. Calculate the mean of each type of government data in different time dimensions for the target type of government data.

[0097] S202. Based on the mean of each type of government data in different time dimensions, calculate the year-on-year growth rate of the mean of the government data in the time dimension.

[0098] In this way, the government data indicators corresponding to each enterprise include: the average amount of VAT payments in the past three months, the average amount of VAT payments in the past six months, the average amount of VAT payments in the past twelve months, the year-on-year growth rate of the average VAT payments in the past three months, the year-on-year growth rate of the average VAT payments in the past six months, the year-on-year growth rate of the average VAT payments in the past twelve months, the average income tax payments in the past three months, the average income tax payments in the past six months, the average income tax payments in the past twelve months, the year-on-year growth rate of the average income tax payments in the past three months, the year-on-year growth rate of the average income tax payments in the past six months, the year-on-year growth rate of the average income tax payments in the past twelve months, the average amount of employee basic pension insurance (unit payment) payments in the past six months, and the average amount of income tax payments in the past six months. The average monthly basic pension insurance (paid by the employer) payment amount for employees, the average basic pension insurance (paid by the employer) payment amount for employees in the past twelve months, the year-on-year growth rate of the average basic pension insurance (paid by the employer) payment amount for employees in the past three months, the year-on-year growth rate of the average basic pension insurance (paid by the employer) payment amount for employees in the past six months, the year-on-year growth rate of the average basic pension insurance (paid by the employer) payment amount for employees in the past twelve months, the average housing provident fund contribution amount in the past three months, the average housing provident fund contribution amount in the past six months, the average housing provident fund contribution amount in the past twelve months, the year-on-year growth rate of the average housing provident fund contribution amount in the past three months, the year-on-year growth rate of the average housing provident fund contribution amount in the past six months, and the year-on-year growth rate of the average housing provident fund contribution amount in the past twelve months, totaling 24 government data indicators.

[0099] That is, the indicators include the mean and the year-on-year growth rate of the mean.

[0100] Here, government data indicators can also be called variables, model input variables, government data variables, etc.

[0101] In the step S104, the access classification model constructed by training is constructed based on the sample benchmark enterprise set and the sample candidate enterprise set until the access classification model meets the preset training stop condition, thereby obtaining a trained access classification model; the trained access classification model outputs the confidence that the first part of the benchmark enterprises and the first candidate enterprise belong to the access category.

[0102] Before training the constructed admission classification model based on the sample benchmark enterprise set and the sample candidate enterprise set, the method further includes:

[0103] Multiple indicators of each type of government data in multiple time dimensions are used as input variables to construct an access classification model based on a binary classification algorithm.

[0104] For example only, the specific process of establishing the admission classification model using the binary classification model of light GBM is as follows:

[0105] First set the default parameter params['learning_rate'] = 0.03#learning rate

[0106] params['boosting_type']='gbdt'#GradientBoostingDecisionTree

[0107] params['objective'] = 'binary' #Select binary classification here

[0108] params['metric'] = 'binary_logloss' # Binary classification loss function

[0109] params['max_depth'] = 8 # Maximum depth of the tree

[0110] params['num_leaves'] = 256

[0111] params['feature_fraction']=0.8

[0112] params['bagging_fraction']=0.8

[0113] params['num_threads'] = 8 #Number of threads. It is recommended that this number be equal to the number of CPU cores.

[0114] params['verbosity'] = 20

[0115] params['early_stopping_round']=100.

[0116] Specifically, a model for identifying small and micro enterprises that have received loans is constructed based on the LightGBM binary classification algorithm.

[0117] Considering that the model effect is to distinguish the enterprises that have obtained loans, the accuracy (Precision) in the traditional binary classification model evaluation method is adopted, and the model parameters are adjusted until the required accuracy is achieved.

[0118] That is, the constructed admission classification model is trained based on the sample benchmark enterprise set and the sample candidate enterprise set until the admission classification model meets the preset training stop condition, thereby obtaining the trained admission classification model, including:

[0119] Based on the access classification model constructed by training the sample benchmark enterprise set and the sample candidate enterprise set, the parameters of the access classification model are adjusted based on the accuracy of the output result of the access classification model until the accuracy of the output result of the access classification model meets the preset accuracy threshold.

[0120] As an example only, the preset accuracy threshold is 90%.

[0121] In the embodiment of the present application, model optimization is not performed based on the discrimination evaluation indicators of the general binary classification model (such as KS, AUC, etc.) because there are no negative samples in the data set described in the embodiment of the present application, and the goal of the model is to screen out more suitable enterprises from the credit-free enterprises. The accuracy rate can better determine the model effect.

[0122] In the step S103, based on the confidence of the first part of benchmarking enterprises and the confidence of the first candidate enterprise, a target enterprise to enter the credit review node is selected from the first candidate enterprises.

[0123] The confidence that the first part of benchmarking enterprises belong to the admission category and the confidence that the first candidate enterprise belongs to the admission category calculated at the end of training of the trained admission classification model are obtained, and the admission threshold of the first candidate enterprise is determined based on the confidence of the first part of benchmarking enterprises.

[0124] The access classification model outputs the confidence level of each enterprise belonging to the access category and the confidence level of each enterprise belonging to the non-access category; the confidence level of each enterprise belonging to the access category represents the probability that the enterprise has good credit.

[0125] According to the admission prediction probability of the confidence level of the benchmark enterprises in the first part belonging to the admission category, its quantile is selected as the admission threshold θ1 (exemplarily, θ1 is 95%), and based on the confidence level in the first candidate enterprise and the admission threshold θ1, the target enterprise to enter the credit review node is selected.

[0126] In some embodiments, see, Figure 3 A flow chart of a method for selecting a target enterprise to enter a credit review node from the first candidate enterprises according to an embodiment of the present application is shown; specifically, selecting a target enterprise to enter a credit review node from the first candidate enterprises according to the confidence of the benchmarking enterprises in the first part and the confidence of the first candidate enterprise includes the following steps S301-S303:

[0127] S301, determining an admission threshold of a first candidate enterprise based on the confidence of the first part of benchmarking enterprises;

[0128] S302: based on the admission threshold, screen out first candidate enterprises to be admitted whose confidence exceeds the admission threshold from the first candidate enterprises at one time;

[0129] S303. A preset number of first candidate enterprises with higher confidence rankings are selected from the first candidate enterprises to be admitted as target enterprises entering the credit review node.

[0130] The determining the admission threshold of the first candidate enterprise based on the confidence of the first part of benchmarking enterprises includes:

[0131] Determine the admission threshold of the first candidate enterprise based on the confidence of the first part of benchmarking enterprises;

[0132] According to the admission target and the number N of the first candidate enterprises, the admission threshold of the first candidate enterprises is adjusted within a preset threshold range to obtain an updated admission threshold.

[0133] Exemplarily, the access target can be determined based on the target average loan amount, the target number of enterprises to be admitted, etc. For example, if the target average loan amount is small, the access threshold is lowered within a preset threshold range; conversely, if the target average loan amount is large, the access threshold is increased within a preset threshold range.

[0134] Specifically, the evaluation result of whether the target loan amount is more or less can be determined based on the comparison result with the standard loan amount.

[0135] Adjusting the admission threshold within the preset threshold range ensures that the credit of the admitted credit-free enterprises is still reliable even after the admission threshold is adjusted.

[0136] A preset number of first candidate enterprises ranked first in confidence are selected from the first candidate enterprises to be admitted as target enterprises entering the credit review node. As an example only, 50% of the enterprises ranked first in admission prediction probability are selected to enter the credit review node.

[0137] In the embodiment of the present application, the approved credit-free enterprises are screened twice. The first screening is based on the admission threshold, which can quickly exclude those enterprises whose predicted admission probability is lower than the threshold, and preliminarily screen out those enterprises that are unlikely to meet the loan conditions from the statistical and risk assessment perspectives, so as to reduce the workload of subsequent detailed evaluation; after the first screening, although a group of credit-free enterprises to be admitted have been obtained, there are still differences in the predicted admission probabilities of these enterprises. The purpose of the second screening is to further select the enterprises with the greatest possibility of successful repayment and the lowest risk from these enterprises to be admitted, and at the same time, the credit-free enterprises to be admitted selected in the first screening will be used as alternative enterprises.

[0138] In the embodiment of the present application, the remaining PS benchmarking enterprises with a label of 1 form a seed sample enterprise set, and the S benchmarking enterprises and N credit-free enterprises form a non-seed sample enterprise set with a label of 0 to train the admission classification model. Not only is the trained admission classification model obtained for subsequent predictions, but also the admitted enterprises among the N credit-free enterprises in this batch are obtained at the same time.

[0139] Get the trained admission classification model for subsequent prediction. For details, please refer to Figure 4 , Figure 4 A flowchart of another method for selecting a target enterprise based on a classification model according to an embodiment of the present application is shown; Figure 4 As shown, after selecting the target enterprise to enter the credit review node from the first candidate enterprises, the method further includes the following steps S401-S404:

[0140] S401. When the target data platform detects a second candidate enterprise that does not meet the preset credit conditions, obtain the credit information of the second candidate enterprise from the target data platform; the credit information of the second candidate enterprise also includes government affairs data of the target type;

[0141] S402, processing the target type of government data in the credit information of the second candidate enterprise, and determining multiple indicators of multiple types of government data of the second candidate enterprise in multiple time dimensions;

[0142] S403, processing multiple indicators of multiple types of government affairs data of the second candidate enterprise in multiple time dimensions through the trained admission classification model, and outputting the confidence that the second candidate enterprise belongs to the admission category;

[0143] S404: Based on the confidence level of the second candidate enterprise, determine whether the second candidate enterprise enters a credit review node.

[0144] The second candidate enterprise is different from the first candidate enterprise, that is, it is a new enterprise with no credit record other than the first candidate enterprise. These new enterprises with no credit record can directly use the trained admission classification model to predict their admission probability and determine their admission results.

[0145] The target data platform detects a second candidate enterprise that does not meet the preset credit conditions. For example, the second candidate enterprise submits a credit review request to the target data platform, and so on.

[0146] Since the target data platform detects that the number of second candidate enterprises that do not meet the preset credit conditions may be relatively small, it can directly determine that when the confidence that the second candidate enterprise belongs to the admission category exceeds the admission threshold, the second candidate enterprise is controlled to automatically enter the credit review node.

[0147] Based on the same inventive concept, the embodiments of the present application also provide a device for selecting target enterprises based on a classification model corresponding to the method for selecting target enterprises based on a classification model. Since the principle of solving the problem by the device in the embodiments of the present application is similar to the method for selecting target enterprises based on a classification model in the embodiments of the present application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0148] Please refer to Figure 5 , Figure 5 The schematic diagram of the structure of the device for selecting a target enterprise based on a classification model according to an embodiment of the present application is shown as follows: Figure 5 As shown, the device comprises:

[0149] The first acquisition module 501 is used to pre-access the target data platform and obtain the credit information of multiple benchmark enterprises that meet the credit acquisition conditions from the target data platform, wherein the credit information of the benchmark enterprises includes the government affairs data of the target type;

[0150] The second acquisition module 502 is used to acquire, from the target data platform, the credit information of a first candidate enterprise whose credit information does not meet the preset credit conditions; the credit information of the first candidate enterprise also includes government affairs data of the target type; the first candidate enterprise is an enterprise to enter the credit review node;

[0151] The sampling module 503 is used to sample and select a first part of benchmarking enterprises from the benchmarking enterprises, wherein the first part of benchmarking enterprises and the first candidate enterprise constitute a sample candidate enterprise set, and the remaining second part of benchmarking enterprises constitute a sample benchmarking enterprise set; the sample benchmarking enterprise set and the sample candidate enterprise set include government affairs data of target types corresponding to the enterprises; the category of the sample candidate enterprise set is an access category;

[0152] A training module 504 is used to train the constructed admission classification model based on the sample benchmark enterprise set and the sample candidate enterprise set until the admission classification model meets a preset training stop condition, thereby obtaining a trained admission classification model; the trained admission classification model outputs the confidence that the first part of the benchmark enterprises and the first candidate enterprise belong to the admission category;

[0153] The selection module 505 is used to select a target enterprise to enter the credit review node from the first candidate enterprises according to the confidence of the first part of benchmark enterprises and the confidence of the first candidate enterprises.

[0154] In some embodiments, the device for selecting a target enterprise based on a classification model further comprises:

[0155] A processing module, used for processing the target type of government data to obtain multiple indicators of each type of government data in multiple time dimensions;

[0156] The target type of government data corresponding to the enterprises in the sample benchmark enterprise set and the sample candidate enterprise set are multiple indicators of each type of government data in multiple time dimensions.

[0157] In some embodiments, the processing module of the device for selecting a target enterprise based on a classification model, when processing the government data of the target type to obtain multiple indicators of each type of government data in multiple time dimensions, is specifically used to:

[0158] For the target type of government affairs data, calculating the mean of each type of government affairs data in different time dimensions;

[0159] Based on the mean of each type of government data in different time dimensions, the year-on-year growth rate of the mean of this type of government data in this time dimension is calculated.

[0160] In some embodiments, the device for selecting a target enterprise based on a classification model further comprises:

[0161] A construction module is used to use multiple indicators of each type of government data in multiple time dimensions as input variables to construct an access classification model based on a binary classification algorithm before training and constructing the access classification model based on the sample benchmark enterprise set and the sample candidate enterprise set.

[0162] In some embodiments, in the device for selecting a target enterprise based on a classification model, the credit acquisition conditions include:

[0163] The type of loan obtained by the benchmark enterprise is a specified type, and the time when the loan is obtained by the benchmark enterprise is within a first preset period, and no specified risk event occurs to the benchmark enterprise within a second preset period.

[0164] In some embodiments, the selection module of the device for selecting a target enterprise based on a classification model, when selecting a target enterprise to enter a credit review node from the first candidate enterprises according to the confidence of the first part of the benchmarking enterprises and the confidence of the first candidate enterprise, is specifically used to:

[0165] Determine the admission threshold of the first candidate enterprise based on the confidence of the first part of benchmarking enterprises;

[0166] Based on the admission threshold, selecting first candidate enterprises to be admitted whose confidence exceeds the admission threshold from the first candidate enterprises at one time;

[0167] A preset number of first candidate enterprises with higher confidence rankings are selected from the first candidate enterprises to be admitted as target enterprises entering the credit review node.

[0168] In some embodiments, the device for selecting a target enterprise based on a classification model further comprises:

[0169] A prediction module, which is used for obtaining the credit information of a second candidate enterprise from the target data platform when the target data platform detects that the second candidate enterprise does not meet the preset credit conditions; the credit information of the second candidate enterprise also includes the government affairs data of the target type;

[0170] Processing the target type of government data in the credit information of the second candidate enterprise, and determining multiple indicators of multiple types of government data of the second candidate enterprise in multiple time dimensions;

[0171] Processing multiple indicators of multiple types of government affairs data of the second candidate enterprise in multiple time dimensions through the trained admission classification model, and outputting the confidence that the second candidate enterprise belongs to the admission category;

[0172] Based on the confidence of the second candidate enterprise, determine whether the second candidate enterprise enters the credit review node.

[0173] Based on the same inventive concept, an electronic device corresponding to the method for selecting target enterprises based on a classification model is also provided in the embodiments of the present application. Since the principle of solving the problem by the electronic device in the embodiments of the present application is similar to the method for selecting target enterprises based on a classification model in the embodiments of the present application, the implementation of the electronic device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0174] Please refer to Figure 6 , Figure 6 A schematic diagram of the structure of an electronic device described in an embodiment of the present application is shown; the electronic device 600 includes: a processor 601, a memory 602 and a bus, the memory 602 stores machine-readable instructions executable by the processor 601, and when the electronic device 600 is running, the processor 601 communicates with the memory 602 through the bus, and when the machine-readable instructions are executed by the processor 601, the steps of the method for selecting a target enterprise based on a classification model are performed.

[0175] Based on the same inventive concept, a computer-readable storage medium corresponding to the method for selecting target enterprises based on a classification model is also provided in the embodiments of the present application. Since the principle of solving the problem by the computer-readable storage medium in the embodiments of the present application is similar to the method for selecting target enterprises based on a classification model in the embodiments of the present application, the implementation of the computer-readable storage medium can refer to the implementation of the method, and the repeated parts will not be repeated.

[0176] A computer-readable storage medium stores a computer program, which, when executed by a processor, executes the steps of the method for selecting a target enterprise based on a classification model.

[0177] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the system and device described above can refer to the corresponding process in the method embodiment, and will not be repeated in this application. In the several embodiments provided in this application, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0178] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0179] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0180] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a platform server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard drives, ROM, RAM, magnetic disks, or optical disks.

[0181] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for selecting target enterprises based on a classification model, characterized in that: Applied to a server or a terminal device, the method comprises: Accessing the target data platform in advance, and obtaining credit information of multiple benchmarking enterprises that meet the credit acquisition conditions from the target data platform, wherein the credit information of the benchmarking enterprises includes government affairs data of the target type; Acquiring from the target data platform the credit information of a first candidate enterprise whose credit information does not meet the preset credit conditions; the credit information of the first candidate enterprise also includes government affairs data of the target type; the first candidate enterprise is an enterprise to enter a credit review node; A first part of benchmarking enterprises are sampled and selected from the benchmarking enterprises, the first part of benchmarking enterprises and the first candidate enterprise constitute a sample candidate enterprise set, and the remaining second part of benchmarking enterprises constitute a sample benchmarking enterprise set; the sample benchmarking enterprise set and the sample candidate enterprise set include government affairs data of target types corresponding to the enterprises; the category of the sample candidate enterprise set is the access category; The access classification model constructed by training based on the sample benchmark enterprise set and the sample candidate enterprise set is obtained until the access classification model meets a preset training stop condition; the trained access classification model outputs the confidence that the first part of the benchmark enterprises and the first candidate enterprise belong to the access category; According to the confidence of the benchmark enterprises in the first part and the confidence of the first candidate enterprise, a target enterprise to enter the credit review node is selected from the first candidate enterprises.

2. The method for selecting target enterprises based on a classification model according to claim 1, characterized in that: The method further comprises: Processing the target type of government data to obtain multiple indicators of each type of government data in multiple time dimensions; The target type of government data corresponding to the enterprises in the sample benchmark enterprise set and the sample candidate enterprise set are multiple indicators of each type of government data in multiple time dimensions.

3. The method for selecting target enterprises based on a classification model according to claim 2, characterized in that: The target type of government data is processed to obtain multiple indicators of each type of government data in multiple time dimensions, including: For the target type of government affairs data, calculating the mean of each type of government affairs data in different time dimensions; Based on the mean of each type of government data in different time dimensions, the year-on-year growth rate of the mean of this type of government data in this time dimension is calculated.

4. The method for selecting target enterprises based on a classification model according to claim 2, characterized in that: Before training the constructed admission classification model based on the sample benchmark enterprise set and the sample candidate enterprise set, the method further includes: Multiple indicators of each type of government data in multiple time dimensions are used as input variables to construct an access classification model based on a binary classification algorithm.

5. The method for selecting target enterprises based on a classification model according to claim 1, characterized in that: The credit acquisition conditions include: The type of loan obtained by the benchmark enterprise is a specified type, and the time when the loan is obtained by the benchmark enterprise is within a first preset period, and no specified risk event occurs to the benchmark enterprise within a second preset period.

6. The method for selecting target enterprises based on a classification model according to claim 4, characterized in that: According to the confidence of the benchmarking enterprise in the first part and the confidence of the first candidate enterprise, a target enterprise to enter the credit review node is selected from the first candidate enterprises, including: Determine the admission threshold of the first candidate enterprise based on the confidence of the first part of benchmarking enterprises; Based on the admission threshold, selecting first candidate enterprises to be admitted whose confidence exceeds the admission threshold from the first candidate enterprises at one time; A preset number of first candidate enterprises with higher confidence rankings are selected from the first candidate enterprises to be admitted as target enterprises entering the credit review node.

7. The method for selecting target enterprises based on a classification model according to claim 1, characterized in that: After selecting a target enterprise to enter a credit review node from the first candidate enterprises, the method further includes: When the target data platform detects a second candidate enterprise that does not meet the preset credit conditions, the credit information of the second candidate enterprise is obtained from the target data platform; the credit information of the second candidate enterprise also includes the government affairs data of the target type; Processing the target type of government data in the credit information of the second candidate enterprise, and determining multiple indicators of multiple types of government data of the second candidate enterprise in multiple time dimensions; Processing multiple indicators of multiple types of government affairs data of the second candidate enterprise in multiple time dimensions through the trained admission classification model, and outputting the confidence that the second candidate enterprise belongs to the admission category; Based on the confidence of the second candidate enterprise, determine whether the second candidate enterprise enters the credit review node.

8. A device for selecting a target enterprise based on a classification model, characterized in that: Applied to a server or terminal device, the device comprises: A first acquisition module is used to pre-access the target data platform and acquire the credit information of a plurality of benchmark enterprises that meet the credit acquisition conditions from the target data platform, wherein the credit information of the benchmark enterprises includes government affairs data of the target type; A second acquisition module is used to acquire, from the target data platform, the credit information of a first candidate enterprise whose credit information does not meet the preset credit conditions; the credit information of the first candidate enterprise also includes government affairs data of the target type; the first candidate enterprise is an enterprise to enter a credit review node; A sampling module is used to sample and select a first part of benchmarking enterprises from the benchmarking enterprises, wherein the first part of benchmarking enterprises and the first candidate enterprise constitute a sample candidate enterprise set, and the remaining second part of benchmarking enterprises constitute a sample benchmarking enterprise set; the sample benchmarking enterprise set and the sample candidate enterprise set include government affairs data of target types corresponding to the enterprises; the category of the sample candidate enterprise set is an access category; A training module is used to train the constructed admission classification model based on the sample benchmark enterprise set and the sample candidate enterprise set until the admission classification model meets a preset training stop condition, thereby obtaining a trained admission classification model; the trained admission classification model outputs the confidence that the first part of the benchmark enterprises and the first candidate enterprise belong to the admission category; A selection module is used to select a target enterprise to enter the credit review node from the first candidate enterprises based on the confidence of the first part of benchmark enterprises and the confidence of the first candidate enterprises.

9. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the method for selecting a target enterprise based on a classification model as described in any one of claims 1 to 7 are performed.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, executes the steps of the method for selecting a target enterprise based on a classification model as described in any one of claims 1 to 7.

Images