Risk assessment method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to a risk assessment method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: in response to a received risk assessment instruction, determining a to-be-assessed organization and a target business type corresponding to the risk assessment instruction, and obtaining to-be-assessed data of the to-be-assessed organization; inputting the to-be-evaluated data into a target evaluation model matched with the target business type, and outputting a predicted risk score through the target evaluation model, the target evaluation model being obtained by training according to a corresponding relationship between training sample data and a corresponding risk score label, the training sample data is obtained after screening by using a preset data screening strategy; and when the predicted risk score is less than a preset threshold value, determining that the to-be-assessed organization passes verification. By adopting the method, the accuracy and the reliability can be improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a risk assessment method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Art

[0002] In order to accurately assess the credit status of an enterprise and avoid potential risks, the enterprise credit-related data can be analyzed to obtain an enterprise credit score that is used to characterize the enterprise's credit status.

[0003] In related technologies, scoring models can be used to analyze corporate credit-related data and output corporate credit scores. However, the model relies on a large amount of high-quality data for training and prediction. In actual applications, corporate credit-related data involves multiple aspects, and there are large differences in data between companies of different industries and sizes. The data quality used for model training is poor, and the model output accuracy is poor, making it difficult to obtain accurate and reliable corporate credit scores. Summary of the invention

[0004] Based on this, it is necessary to provide a risk assessment method, apparatus, computer equipment, computer-readable storage medium and computer program product that can improve accuracy and reliability in response to the above technical problems.

[0005] In a first aspect, the present application provides a risk assessment method, comprising:

[0006] In response to receiving a risk assessment instruction, determining an organization to be assessed and a target business type corresponding to the risk assessment instruction, and obtaining data to be assessed of the organization to be assessed;

[0007] Input the data to be evaluated into a target evaluation model that matches the target business type, and output a predicted risk score through the target evaluation model, wherein the target evaluation model is trained according to the correspondence between the training sample data and the corresponding risk score label, and the training sample data is obtained after screening using a preset data screening strategy;

[0008] When the predicted risk score is less than a preset threshold, it is determined that the tissue to be evaluated has passed the verification.

[0009] In one embodiment, the target assessment model is determined by:

[0010] Acquire a training sample data set, wherein the training sample data set includes training sample data annotated with a risk score label;

[0011] Constructing an initial evaluation model, wherein the initial evaluation model is provided with training parameters;

[0012] Inputting the training sample data set into the initial evaluation model, and obtaining an output result through the initial evaluation model;

[0013] The training parameters are adjusted according to the difference between the output result and the corresponding risk score label until the difference meets the preset requirements to obtain the target evaluation model.

[0014] In one embodiment, the method for obtaining the training sample data includes:

[0015] Obtain the initial organization-related data corresponding to the target business type;

[0016] Dividing the initial organization-related data according to a preset data dimension to obtain divided data, wherein the divided data includes at least one of organization data, resource data, organization honor data, business data, and organization information base data;

[0017] Processing the divided data according to a preset data screening strategy to obtain screened data;

[0018] The risk score of the filtered data is determined according to the risk data of the organization corresponding to the filtered data, and the risk score is used as a label to annotate the filtered data to obtain training sample data.

[0019] In one embodiment, the processing of the divided data according to a preset data screening strategy includes:

[0020] Obtaining a preset data screening strategy, wherein the preset data screening strategy includes data screening rules and corresponding data ranges, and the data screening rules include at least one of data compliance rules, data proportion rules, data semantic rules, data complexity rules, and data distribution rules;

[0021] The divided data are screened according to the data screening rule to obtain the screened data within the data range.

[0022] In one embodiment, until the difference meets the preset requirements, a target evaluation model is obtained, including:

[0023] Until the difference meets the preset requirements, an intermediate evaluation model is obtained;

[0024] Inputting the training sample data set into the intermediate evaluation model, and obtaining an output result through the intermediate evaluation model;

[0025] When the distribution of the output results meets a preset condition, the intermediate evaluation model is determined to be a target evaluation model.

[0026] In one of the embodiments, when the distribution of the output results meets a preset condition, determining the intermediate evaluation model as the target evaluation model includes:

[0027] Dividing the output results according to the organizational type of the organization of the training sample data corresponding to the output results to obtain multiple groups of output results;

[0028] Obtaining risk distribution information corresponding to the organization type;

[0029] For each group of the output results, comparing the distribution information of the output results and the corresponding risk distribution information, and determining that the output results pass the verification when the difference between the distribution information and the risk distribution information is less than a preset difference value;

[0030] When the plurality of groups of output results all pass verification, the intermediate evaluation model is determined as the target evaluation model.

[0031] In a second aspect, the present application also provides a risk assessment device, comprising:

[0032] A response module, configured to, in response to receiving a risk assessment instruction, determine the organization to be assessed and the target business type corresponding to the risk assessment instruction, and obtain the data to be assessed of the organization to be assessed;

[0033] An input module, used for inputting the data to be evaluated into a target evaluation model matching the target business type, and outputting a predicted risk score through the target evaluation model, wherein the target evaluation model is trained according to the correspondence between the training sample data and the corresponding risk score label, and the training sample data is obtained after screening using a preset data screening strategy;

[0034] The determination module is used to determine that the organization to be evaluated has passed the verification when the predicted risk score is less than a preset threshold.

[0035] In one embodiment, the device further includes a determination module of the target assessment model, and the determination module of the target assessment model includes:

[0036] A first acquisition submodule is used to acquire a training sample data set, wherein the training sample data set includes training sample data marked with a risk score label;

[0037] A construction submodule is used to construct an initial evaluation model, wherein the initial evaluation model is provided with training parameters;

[0038] An input submodule, used for inputting the training sample data set into the initial evaluation model, and obtaining an output result through the initial evaluation model;

[0039] The adjustment submodule is used to adjust the training parameters according to the difference between the output result and the corresponding risk score label until the difference meets the preset requirements to obtain the target evaluation model.

[0040] In one embodiment, the device further includes a module for acquiring the training sample data, and the module for acquiring the training sample data includes:

[0041] The second acquisition submodule is used to acquire the initial organization-related data corresponding to the target business type;

[0042] A division submodule, used to divide the initial organization-related data according to a preset data dimension to obtain divided data, wherein the divided data includes at least one of organization data, resource data, organization honor data, business data, and organization information base data;

[0043] A screening submodule, used to process the divided data according to a preset data screening strategy to obtain screened data;

[0044] A determination submodule is used to determine the risk score of the screened data according to the risk data of the organization corresponding to the screened data, and use the risk score as a label to mark the screened data to obtain training sample data.

[0045] In one embodiment, the screening submodule is further used for:

[0046] Obtaining a preset data screening strategy, wherein the preset data screening strategy includes data screening rules and corresponding data ranges, and the data screening rules include at least one of data compliance rules, data proportion rules, data semantic rules, data complexity rules, and data distribution rules;

[0047] The divided data are screened according to the data screening rule to obtain the screened data within the data range.

[0048] In one embodiment, the adjustment submodule is further used for:

[0049] Until the difference meets the preset requirements, an intermediate evaluation model is obtained;

[0050] Inputting the training sample data set into the intermediate evaluation model, and obtaining an output result through the intermediate evaluation model;

[0051] When the distribution of the output results meets a preset condition, the intermediate evaluation model is determined to be a target evaluation model.

[0052] In one embodiment, the adjustment submodule is further used for:

[0053] Dividing the output results according to the organizational type of the organization of the training sample data corresponding to the output results to obtain multiple groups of output results;

[0054] Obtaining risk distribution information corresponding to the organization type;

[0055] For each group of the output results, comparing the distribution information of the output results and the corresponding risk distribution information, and determining that the output results pass the verification when the difference between the distribution information and the risk distribution information is less than a preset difference value;

[0056] When the plurality of groups of output results all pass verification, the intermediate evaluation model is determined as the target evaluation model.

[0057] In a third aspect, an embodiment of the present disclosure further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of any one of the methods in the embodiments of the present disclosure are implemented.

[0058] In a fourth aspect, the embodiments of the present disclosure further provide a computer-readable storage medium, wherein a computer program is stored thereon, and when the computer program is executed by a processor, the steps of any one of the methods in the embodiments of the present disclosure are implemented.

[0059] In a fifth aspect, the embodiments of the present disclosure further provide a computer program product, wherein the computer program product includes a computer program, and when the computer program is executed by a processor, the steps of any one of the methods in the embodiments of the present disclosure are implemented.

[0060] The above-mentioned risk assessment method, device, computer equipment, computer-readable storage medium and computer program product, after receiving the risk assessment instruction, determine the organization to be assessed and the target business type corresponding to the risk assessment instruction, and obtain the data to be assessed of the organization to be assessed, input the data to be assessed into the target assessment model that matches the target business type, and output the predicted risk score through the target assessment model. When the predicted risk score is less than the preset threshold, it is determined that the organization to be assessed has passed the verification, so that the risk assessment of the organization to be assessed can be performed quickly and accurately, potential risks can be avoided, and the security and reliability of the business can be ensured. Different business types correspond to different models, so that risk assessment can be carried out according to different businesses in a targeted manner, the flexibility and practicality of risk assessment can be improved, the efficiency and automation of risk assessment can be improved, and the accuracy of predicted risk scores can be ensured. The target assessment model is trained using the corresponding relationship between the training sample data and the corresponding risk score label, and the training sample data is obtained after screening using the preset data screening strategy. Data screening using the preset data screening strategy can effectively improve the data quality of the sample data used for model training, improve the output accuracy of the trained model, thereby obtaining an accurate and reliable predicted risk score, effectively conducting risk assessment on the organization to be assessed, reducing the possibility of potential risks, and being applicable to more application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0062] Figure 1 A schematic diagram of a risk assessment method in one embodiment;

[0063] Figure 2 A schematic diagram of a flow chart of a method for determining a target evaluation model in one embodiment;

[0064] Figure 3 is a flow chart of a method for acquiring training sample data in one embodiment;

[0065] Figure 4 is a structural block diagram of a risk assessment device in one embodiment;

[0066] Figure 5 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0068] In one embodiment, Figure 1 As shown, a risk assessment method is provided. This embodiment uses the method applied to a terminal as an example. It can be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0069] Step S110, in response to receiving the risk assessment instruction, determining the organization to be assessed and the target business type corresponding to the risk assessment instruction, and obtaining the data to be assessed of the organization to be assessed;

[0070] Exemplarily, in response to receiving a risk assessment instruction, the organization to be assessed and the target business type corresponding to the risk assessment instruction are determined. For different business types, the data dimensions and assessment rules to be assessed may be different. Therefore, when performing a risk assessment, the organization to be assessed and the target business type need to be determined according to the risk assessment instruction.

[0071] Optionally, the data to be evaluated of the organization to be evaluated is obtained, wherein the data to be evaluated matches the target business type, and different target business types may correspond to data to be evaluated of different data dimensions.

[0072] Step S120, inputting the data to be evaluated into a target evaluation model that matches the target business type, and outputting a predicted risk score through the target evaluation model, wherein the target evaluation model is trained according to the correspondence between the training sample data and the corresponding risk score label, and the training sample data is obtained after screening using a preset data screening strategy;

[0073] Exemplarily, a target assessment model that matches the target business type is determined. In one example, different business types correspond to different assessment models. The data to be assessed is input into the target assessment model, and the target assessment model outputs a predicted risk score.

[0074] Optionally, the target assessment model is trained based on the correspondence between the training sample data and the corresponding risk score label, and the training sample data is obtained after screening using a preset data screening strategy. In one example, the training sample data can be determined based on historical data of different organizations. The preset data screening strategy can be determined based on data screening rules associated with data quality in actual application scenarios.

[0075] In one example, during the training of the target assessment model, the training sample data corresponding to the organizations with risks are recorded as bad samples, and the training sample data corresponding to the organizations without risks are recorded as good samples. The model outputs the predicted bad sample ratio, which corresponds to the predicted risk score. In one example, for data with characteristics such as not included in the model, insufficient weight, and small scenarios, relevant policies and rules are specified to adjust the score.

[0076] Step S130: When the predicted risk score is less than a preset threshold, determining that the tissue to be evaluated has passed the verification.

[0077] For example, a higher predicted risk score indicates a greater possibility that the organization to be assessed has a risk, and a lower predicted risk score indicates a lower possibility that the organization to be assessed has a risk. The organization to be assessed that is determined to have a lower possibility of risk based on the predicted risk score passes the verification.

[0078] In one example, a preset threshold is set according to an actual application scenario. When the predicted risk score is less than the preset threshold, it can be considered that the possibility of risk in the organization to be assessed is small, and the business request corresponding to the target business type can be approved. Therefore, it is determined that the organization to be assessed has passed the verification.

[0079] Optionally, when the predicted risk score is greater than or equal to the preset threshold, it can be considered that the organization to be assessed is more likely to have risks. Therefore, it is determined that the organization to be assessed has failed the verification and cannot pass the business request corresponding to the target business type.

[0080] In the embodiment of the present disclosure, after receiving the risk assessment instruction, the organization to be assessed and the target business type corresponding to the risk assessment instruction are determined, and the data to be assessed of the organization to be assessed are obtained, and the data to be assessed is input into the target assessment model that matches the target business type, and the target assessment model outputs the predicted risk score. When the predicted risk score is less than the preset threshold, it is determined that the organization to be assessed has passed the verification, so that the risk assessment of the organization to be assessed can be performed quickly and accurately, potential risks can be avoided, and the security and reliability of the business can be ensured. Different business types correspond to different models, so that risk assessment can be performed in a targeted manner according to different businesses, the flexibility and practicality of risk assessment are improved, the efficiency and automation of risk assessment are improved, and the accuracy of predicted risk scores is ensured. The target assessment model is trained using the corresponding relationship between the training sample data and the corresponding risk score label, and the training sample data is obtained after screening using the preset data screening strategy. Data screening using the preset data screening strategy can effectively improve the data quality of the sample data used for model training, and improve the output accuracy of the trained model, so as to obtain accurate and reliable predicted risk scores, conduct effective risk assessment on the organization to be assessed, reduce the possibility of potential risks, and be applicable to more application scenarios.

[0081] In one embodiment, Figure 2 As shown, the method for determining the target evaluation model includes:

[0082] Step S210, obtaining a training sample data set, wherein the training sample data set includes training sample data annotated with a risk score label;

[0083] Step S220, constructing an initial evaluation model, wherein the initial evaluation model is provided with training parameters;

[0084] Step S230, inputting the training sample data set into the initial evaluation model, and obtaining an output result through the initial evaluation model;

[0085] Step S240, adjusting the training parameters according to the difference between the output result and the corresponding risk score label until the difference meets the preset requirements, thereby obtaining a target assessment model.

[0086] Exemplarily, a training sample data set is obtained, wherein the training sample data set includes training sample data labeled with risk score labels. In one example, the training sample data can be determined based on relevant data of multiple different organizations.

[0087] In one example, an initial evaluation model is constructed, and training parameters are set in the initial evaluation model. In a possible implementation, a logistic regression algorithm can be used as a basic algorithm to construct the initial evaluation model. The training parameters can be determined according to the actual application scenario, and different application scenarios can correspond to different training parameters.

[0088] Optionally, the training sample data set is input into the initial evaluation model, and the output result is obtained through the initial evaluation model. Since the initial evaluation model is an untrained model, the model output accuracy is poor and there is a large difference between the output result and the corresponding risk score label. Therefore, the training parameters are adjusted according to the difference between the output result and the corresponding risk score label until the difference between the output result of the model and the corresponding risk score label meets the preset requirements, thereby obtaining the target evaluation model.

[0089] In one example, the preset requirement may be determined according to an actual application scenario. For example, the preset requirement may be set according to the requirement for the accuracy of the model output in the actual application scenario, such as the difference between the model output result and the corresponding risk score being within a preset range.

[0090] When the difference between the output result and the corresponding risk score label meets the preset requirements, it can be considered that the model output accuracy is high at this time, and it can be used for risk assessment, and the model at this time is determined to be the target assessment model. In one example, when the difference between the output result and the corresponding risk score label does not meet the preset requirements, it can be considered that the model output accuracy is low at this time and does not meet the requirements. Therefore, the training parameters are adjusted according to the difference until they meet the preset requirements. In one example, the adjustment can be made through the loss function.

[0091] In the embodiment of the present disclosure, when determining the target evaluation model, a training sample data set is obtained, an initial evaluation model is constructed, the training sample data set is input into the initial evaluation model, an output result is obtained through the initial evaluation model, and the training parameters are adjusted according to the difference between the output result and the corresponding risk score label until the difference meets the preset requirements, thereby obtaining the target evaluation model. In this way, the target evaluation model can be quickly and accurately trained according to the training sample data and the corresponding risk score label, thereby ensuring the accuracy and reliability of the output of the target evaluation model.

[0092] In one embodiment, Figure 3 As shown, the method for obtaining the training sample data includes:

[0093] Step S310, obtaining initial organization-related data corresponding to the target business type;

[0094] Step S320, dividing the initial organization-related data according to a preset data dimension to obtain divided data, wherein the divided data includes at least one of organization data, resource data, organization honor data, business data, and organization information base data;

[0095] Step S330, processing the divided data according to a preset data screening strategy to obtain screened data;

[0096] Step S340, determining the risk score of the screened data according to the risk data of the organization corresponding to the screened data, and using the risk score as a label to annotate the screened data to obtain training sample data.

[0097] Exemplarily, the training sample data is obtained after division and screening. The initial organization-related data corresponding to the target business type is obtained. Among them, different business types can correspond to different types of data, which can be determined according to the actual application scenario. The initial organization-related data is divided according to the preset data dimensions to obtain the divided data. In an example, the preset data dimensions can be determined according to the actual application scenario. For example, the preset data dimensions can include but are not limited to organizational dimensions, resource dimensions, organizational honor dimensions, business dimensions, organizational information library dimensions, etc. The data can be divided into multiple categories of data through the preset data dimensions.

[0098] In one possible implementation, the data of the organizational dimension include organizational background data, such as basic information of the organization, completeness of basic information, financing situation, legal representative situation, etc.; the data of the resource dimension include financial status data, such as basic financial subjects of the industrial and commercial annual report, debt repayment ability, profitability, operating ability, etc.; the data of the organizational honor dimension may include innovation capability data, such as intellectual property rights, rankings, honors, scientific and technological achievements, etc.; the data of the operating dimension may include operating capability data, such as qualification certificates, bidding situation, credit evaluation, market competitiveness, etc.; the data of the organizational information database dimension may include risk status data, such as judicial risk, regulatory risk, compliance risk, operating risk, basic financial risk indicators, stock market risk, etc.

[0099] In one example, corresponding indicators can be created according to the above dimensions, and data cleaning can be performed. Optionally, the indicators and data are verified by dimensions, and the verification dimensions include compliance, sortability, interpretability, diversity and complexity, overall distribution concentration, group differentiation, etc., among which compliance means that the indicators, rules and data meet compliance and subsequent availability, and there are no discriminatory indicators for consumer rights and controversial content. Sorting means that the score has a relatively good sorting ability, the trend of the bad sample ratio corresponding to the high and low segments is monotonous, and the bad sample ratio that can be identified in the low segment is high and does not jump. Interpretability means that the model variables, policy rules, and variable bins all have good business meanings and meet business expectations, the variable trends are consistent with business understanding, and the overall model has a certain interpretable mechanism. Diversity and complexity means that the complexity of the model variables, variable bins, and the overall model is balanced. Overall distribution concentration means that it has a certain degree of decentralization. The ideal distribution is a distribution trend that is high in the middle and low on both sides and close to normal, avoiding abnormal concentration on certain scores, which causes a high probability of rating jumps. The clustering differentiation is to verify the abnormal fluctuations by group (industry, region, enterprise type, enterprise status, enterprise scale, and others).

[0100] For example, in the sample selection process, bad samples are screened out through rules, and all samples except bad samples are good samples. Sampling is considered when the amount of data is too large. In one example, bad samples are defined as the following situations: the enterprise (i.e., organization) has a survival risk: the enterprise is revoked, in liquidation, liquidation filing, bankruptcy liquidation, deregistration, suspension of business; serious judicial risk: dishonest person subject to execution; serious tax dishonesty: major tax violations; serious industrial and commercial dishonesty: serious violations; bond default: external rating is C or below or bond default; extremely high risk: the registered capital reduction ratio is >= 90% and there is a change in shareholders or legal representative; extremely high risk: the enterprise is involved in a private lending dispute (defendant, appellant).

[0101] In one example, the divided data is processed according to a preset data screening strategy to obtain screened data. Optionally, different categories of data may correspond to different data screening strategies. After processing and screening different categories of data according to different data screening strategies, the screened data is obtained.

[0102] Optionally, a risk score of the filtered data is determined based on the risk data of the organization corresponding to the filtered data, and the risk score is used as a label to annotate the filtered data to obtain training sample data. In one example, the risk score of the data is determined based on the risk situation of the organization corresponding to the data, and the higher the risk, the higher the risk score.

[0103] In one example, two sets of strategy systems, black strategy and green strategy, were established outside the model. Black strategy: mainly a capping strategy, which solves the problem of insufficient strength of the model of the evaluated entity and the non-inclusion of negative data scenarios; Green strategy: a bottoming strategy and a bonus strategy, which solves the problem of bottoming and bonusing the scores of good companies recognized by the public in small scenario data; through monitoring the single hit rate and compound hit rate of black strategy and green strategy, the single and compound hit rates basically meet the overall logic when the rules are formulated, and there should be no significant abnormalities.

[0104] In the disclosed embodiment, when obtaining training sample data, initial organization-related data corresponding to the target business type is obtained, and the initial organization-related data is divided according to preset data dimensions to obtain divided data, and the divided data is processed according to a preset data screening strategy to obtain screened data, and a risk score of the screened data is determined based on the risk data of the organization corresponding to the screened data, and the risk score is used as a label to annotate the screened data to obtain training sample data, so that corresponding relevant data can be obtained based on different business types, and the data can be divided into data of multiple dimensions for processing according to different data dimensions, thereby ensuring the validity and reliability of the data, and processing the divided data using a preset data screening strategy further improves the data quality and ensures the accuracy of the model output.

[0105] In one embodiment, the processing of the divided data according to a preset data screening strategy includes:

[0106] Obtaining a preset data screening strategy, wherein the preset data screening strategy includes data screening rules and corresponding data ranges, and the data screening rules include at least one of data compliance rules, data proportion rules, data semantic rules, data complexity rules, and data distribution rules;

[0107] The divided data are screened according to the data screening rule to obtain the screened data within the data range.

[0108] Exemplarily, the preset screening strategy includes data screening rules and corresponding data ranges, the data screening rules and the data ranges correspond to each other, and different data screening rules correspond to different data ranges. In one example, the data screening rules include at least one of data compliance rules, data proportion rules, data semantic rules, data complexity rules, and data distribution rules.

[0109] The divided data is screened according to the data screening rules to obtain the screened data within the data range. In one example, when the data screening rules include multiple types, the divided data is screened according to the multiple data screening rules in sequence to obtain the final screened data.

[0110] In the disclosed embodiments, when performing data screening, the data is screened according to different data screening rules. Different data screening rules correspond to different data ranges. For different data screening rules, screened data within the corresponding data range is obtained, thereby ensuring that the screened data complies with the data screening rules, improving the accuracy and reliability of the screened data, and being suitable for more application scenarios.

[0111] In one embodiment, until the difference meets the preset requirements, a target evaluation model is obtained, including:

[0112] Until the difference meets the preset requirements, an intermediate evaluation model is obtained;

[0113] Inputting the training sample data set into the intermediate evaluation model, and obtaining an output result through the intermediate evaluation model;

[0114] When the distribution of the output results meets a preset condition, the intermediate evaluation model is determined to be a target evaluation model.

[0115] For example, during the model training process, the distribution of the model output results is verified. When the difference between the model output results and the corresponding risk score labels meets the preset requirements, it can be considered that the model output accuracy meets the requirements, and the model at this time is determined to be an intermediate evaluation model, and the distribution of the model output results is further verified.

[0116] Optionally, the training sample data set is input into the intermediate evaluation model, and the output result is obtained through the intermediate evaluation model. When the distribution of the output result meets the preset conditions, the intermediate evaluation model is determined as the target evaluation model. In one example, the preset conditions can be determined according to the actual application scenario. For example, the preset conditions are set to the distribution of the output result close to the normal distribution, etc., to avoid abnormal concentration of the output results on individual scores, resulting in a high probability of jump in the risk rating.

[0117] In an example, relevant data of some organizations can be sampled as samples for judgment. Most samples are consistent with the auditor's judgment and understanding of credit. If a small number of samples have deviations, the overall explanation can be based on the model's input variables.

[0118] In the embodiment of the present disclosure, when the difference between the model output and the risk score label meets the preset requirements, the distribution of the model output results is verified. When the distribution of the output results meets the preset conditions, the intermediate evaluation model is determined to be the target evaluation model, so that the effect of the model can be verified through the distribution of the model output results, thereby further improving the accuracy and reliability of the model output.

[0119] In one embodiment, when the distribution of the output results meets a preset condition, determining the intermediate evaluation model as a target evaluation model includes:

[0120] Dividing the output results according to the organizational type of the organization of the training sample data corresponding to the output results to obtain multiple groups of output results;

[0121] Obtaining risk distribution information corresponding to the organization type;

[0122] For each group of the output results, comparing the distribution information of the output results and the corresponding risk distribution information, and determining that the output results pass the verification when the difference between the distribution information and the risk distribution information is less than a preset difference value;

[0123] When the plurality of groups of output results all pass verification, the intermediate evaluation model is determined as the target evaluation model.

[0124] Exemplarily, the output results are divided according to the organizational type of the training sample data corresponding to the output results to obtain multiple groups of output results. In one example, the data corresponding to organizations of different organizational types have different distribution characteristics. Therefore, after the output results are divided, the output results can be verified according to the distribution characteristics of the data corresponding to organizations of different organizational types.

[0125] For each tissue type, obtain the risk distribution information corresponding to the tissue type. Determine the distribution information of the output result corresponding to the tissue type. The distribution information of the output result can be considered as the risk distribution information predicted by the model. Compare the distribution information of the output result with the risk distribution information of the corresponding tissue type. If the difference is small, the verification is passed. If the difference is large, the verification is not passed.

[0126] Optionally, a preset difference value is set according to the actual application scenario. When the difference between the distribution information and the corresponding risk distribution information is less than the preset difference value, it can be considered that the distribution of the model output results for the organization type conforms to the risk distribution characteristics of the organization type.

[0127] In one example, when multiple groups of output results all pass verification, the intermediate assessment model is determined to be the target assessment model. Wherein, when multiple groups of output results all pass verification, it can be considered that the predicted risk scores of each tissue type output by the model all meet the corresponding risk distribution characteristics, the model output reliability is high, the model can be used for risk scoring, and the model is determined to be the target assessment model.

[0128] In the disclosed embodiment, when verifying the distribution of the model output results, the output results are divided according to the organizational type of the organization of the training sample data corresponding to the output results to obtain multiple groups of output results, and the distribution message of each group of output results is compared according to the risk distribution message corresponding to the organizational type to determine whether the difference between the distribution information and the risk distribution information is less than a preset difference value. When it is less than the preset difference value, it is determined that the output result has passed the verification. When multiple groups of output results have passed the verification, the intermediate evaluation model is determined to be the target evaluation model, so that a model in which the distribution of the model output results conforms to the actual application scenario can be obtained, thereby ensuring the accuracy and reliability of the model output.

[0129] In some embodiments, when risk assessment is required, an indicator system is built, corresponding single indicators are created based on different data dimensions of the organization, and data cleaning is performed. The indicators are dimensionally verified according to multiple verification dimensions. Two sets of policy systems, black strategy and green strategy, are established outside the model to verify the policy rule system. Model distribution data is verified according to the distribution of model output results. Model evaluation verification and assessment are performed from the business and algorithm perspectives. In one example, a preset type of organization list can be used to perform database collision analysis and calibrate the model. In one example, the model can be deployed on a preset system. In one example, the model can be continuously monitored, and anomaly detection and analysis can be performed regularly and irregularly.

[0130] Through this embodiment, the comprehensiveness and accuracy of predictive risk scoring are improved; the flexibility and adaptability of credit assessment are improved; the application scenarios of credit assessment are effectively expanded; the efficiency and automation of credit assessment are improved; risk assessment is performed according to different business types in multiple business scenarios (access, credit, risk control, etc.), and different risk scores are used as organizational metrics. Scoring and rating can quickly help users understand the overall risk profile of the organization and then identify risk event points.

[0131] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0132] Based on the same inventive concept, the embodiment of the present application also provides a risk assessment device for implementing the risk assessment method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more risk assessment device embodiments provided below can refer to the limitations on the risk assessment method above, and will not be repeated here.

[0133] In an exemplary embodiment, Figure 4 As shown, a risk assessment device 400 is provided, comprising:

[0134] A response module 410 is used to, in response to receiving a risk assessment instruction, determine the organization to be assessed and the target business type corresponding to the risk assessment instruction, and obtain the data to be assessed of the organization to be assessed;

[0135] An input module 420 is used to input the data to be evaluated into a target evaluation model that matches the target business type, and output a predicted risk score through the target evaluation model, wherein the target evaluation model is trained according to the correspondence between the training sample data and the corresponding risk score label, and the training sample data is obtained after screening using a preset data screening strategy;

[0136] The determination module 430 is used to determine that the tissue to be evaluated has passed the verification when the predicted risk score is less than a preset threshold.

[0137] In one embodiment, the device further includes a determination module of the target assessment model, and the determination module of the target assessment model includes:

[0138] A first acquisition submodule is used to acquire a training sample data set, wherein the training sample data set includes training sample data marked with a risk score label;

[0139] A construction submodule is used to construct an initial evaluation model, wherein the initial evaluation model is provided with training parameters;

[0140] An input submodule, used for inputting the training sample data set into the initial evaluation model, and obtaining an output result through the initial evaluation model;

[0141] The adjustment submodule is used to adjust the training parameters according to the difference between the output result and the corresponding risk score label until the difference meets the preset requirements to obtain the target evaluation model.

[0142] In one embodiment, the device further includes a module for acquiring the training sample data, and the module for acquiring the training sample data includes:

[0143] The second acquisition submodule is used to acquire the initial organization-related data corresponding to the target business type;

[0144] A division submodule, used to divide the initial organization-related data according to a preset data dimension to obtain divided data, wherein the divided data includes at least one of organization data, resource data, organization honor data, business data, and organization information base data;

[0145] A screening submodule, used to process the divided data according to a preset data screening strategy to obtain screened data;

[0146] A determination submodule is used to determine the risk score of the screened data according to the risk data of the organization corresponding to the screened data, and to label the screened data with the risk score as a label to obtain training sample data.

[0147] In one embodiment, the screening submodule is further used for:

[0148] Obtaining a preset data screening strategy, wherein the preset data screening strategy includes data screening rules and corresponding data ranges, and the data screening rules include at least one of data compliance rules, data proportion rules, data semantic rules, data complexity rules, and data distribution rules;

[0149] The divided data are screened according to the data screening rule to obtain the screened data within the data range.

[0150] In one embodiment, the adjustment submodule is further used for:

[0151] Until the difference meets the preset requirements, an intermediate evaluation model is obtained;

[0152] Inputting the training sample data set into the intermediate evaluation model, and obtaining an output result through the intermediate evaluation model;

[0153] When the distribution of the output results meets a preset condition, the intermediate evaluation model is determined to be a target evaluation model.

[0154] In one embodiment, the adjustment submodule is further used for:

[0155] Dividing the output results according to the organizational type of the organization of the training sample data corresponding to the output results to obtain multiple groups of output results;

[0156] Obtaining risk distribution information corresponding to the organization type;

[0157] For each group of the output results, comparing the distribution information of the output results and the corresponding risk distribution information, and determining that the output results pass the verification when the difference between the distribution information and the risk distribution information is less than a preset difference value;

[0158] When the plurality of groups of output results all pass verification, the intermediate evaluation model is determined as the target evaluation model.

[0159] Each module in the above risk assessment device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0160] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Figure 5 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data to be evaluated and other data involved in the method described in this embodiment. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a risk assessment method is implemented.

[0161] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0162] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0163] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0164] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0165] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0166] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0167] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0168] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A risk assessment method, characterized in that: The method comprises: In response to receiving a risk assessment instruction, determining an organization to be assessed and a target business type corresponding to the risk assessment instruction, and obtaining data to be assessed of the organization to be assessed; Input the data to be evaluated into a target evaluation model that matches the target business type, and output a predicted risk score through the target evaluation model, wherein the target evaluation model is trained according to the correspondence between the training sample data and the corresponding risk score label, and the training sample data is obtained after screening using a preset data screening strategy; When the predicted risk score is less than a preset threshold, it is determined that the tissue to be evaluated has passed the verification.

2. The method according to claim 1, characterized in that The target evaluation model is determined by: Acquire a training sample data set, wherein the training sample data set includes training sample data annotated with a risk score label; Constructing an initial evaluation model, wherein the initial evaluation model is provided with training parameters; Inputting the training sample data set into the initial evaluation model, and obtaining an output result through the initial evaluation model; The training parameters are adjusted according to the difference between the output result and the corresponding risk score label until the difference meets the preset requirements to obtain the target evaluation model.

3. The method according to claim 2, characterized in that The method for obtaining the training sample data includes: Obtain the initial organization-related data corresponding to the target business type; Dividing the initial organization-related data according to a preset data dimension to obtain divided data, wherein the divided data includes at least one of organization data, resource data, organization honor data, business data, and organization information base data; Processing the divided data according to a preset data screening strategy to obtain screened data; The risk score of the filtered data is determined according to the risk data of the organization corresponding to the filtered data, and the risk score is used as a label to annotate the filtered data to obtain training sample data.

4. The method according to claim 3, characterized in that The processing of the divided data according to the preset data screening strategy includes: Obtaining a preset data screening strategy, wherein the preset data screening strategy includes data screening rules and corresponding data ranges, and the data screening rules include at least one of data compliance rules, data proportion rules, data semantic rules, data complexity rules, and data distribution rules; The divided data are screened according to the data screening rule to obtain the screened data within the data range.

5. The method according to claim 2, characterized in that: Until the difference meets the preset requirements, a target evaluation model is obtained, including: Until the difference meets the preset requirements, an intermediate evaluation model is obtained; Inputting the training sample data set into the intermediate evaluation model, and obtaining an output result through the intermediate evaluation model; When the distribution of the output results meets a preset condition, the intermediate evaluation model is determined to be a target evaluation model.

6. The method according to claim 5, characterized in that When the distribution of the output results meets a preset condition, determining the intermediate evaluation model as a target evaluation model includes: Dividing the output results according to the organizational type of the organization of the training sample data corresponding to the output results to obtain multiple groups of output results; Obtaining risk distribution information corresponding to the organization type; For each group of the output results, comparing the distribution information of the output results and the corresponding risk distribution information, and determining that the output results pass the verification when the difference between the distribution information and the risk distribution information is less than a preset difference value; When the plurality of groups of output results all pass verification, the intermediate evaluation model is determined as the target evaluation model.

7. A risk assessment device, characterized in that: The device comprises: A response module, configured to, in response to receiving a risk assessment instruction, determine the organization to be assessed and the target business type corresponding to the risk assessment instruction, and obtain the data to be assessed of the organization to be assessed; An input module, used for inputting the data to be evaluated into a target evaluation model matching the target business type, and outputting a predicted risk score through the target evaluation model, wherein the target evaluation model is trained according to the correspondence between the training sample data and the corresponding risk score label, and the training sample data is obtained after screening using a preset data screening strategy; The determination module is used to determine that the organization to be evaluated has passed the verification when the predicted risk score is less than a preset threshold.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images