Risk assessment method and device, electronic equipment and storage medium

Abstract

The application discloses a risk assessment method and device, electronic equipment and a storage medium. The risk assessment method comprises the following steps: obtaining a sample to be evaluated; inputting the sample to be evaluated into a first risk model to obtain an initial risk assessment result of the sample to be evaluated; if it is determined based on the initial risk assessment result that the sample to be evaluated belongs to a high-score customer group, inputting the sample to be evaluated into a second risk model to obtain a target risk assessment result of the sample to be evaluated, wherein the second risk model is trained based on target training samples belonging to the high-score customer group in training samples of the first risk model. The application solves the technical problem of poor risk assessment effect in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence technology in financial technology (Fintech), and in particular to a risk assessment method and device, an electronic device and a storage medium. BACKGROUND

[0002] With the continuous development of financial technology, especially Internet technology finance, more and more technologies (such as distributed technology and artificial intelligence) are applied in the financial field, but the financial industry also puts forward higher requirements for technology, such as higher requirements for the distribution of the to-do list of the financial industry.

[0003] In the financial industry, customers are often assessed for risk, and the risk assessment results are then used for credit granting. Generally speaking, the higher the risk of a customer determined by the assessment, the lower the credit limit granted to the customer, and the lower the risk of a customer determined by the assessment, the higher the credit limit granted to the customer. The risk model accurately identifies most high-risk customers, and for those high-risk customers accurately identified, the bank can reject or grant a low credit limit to reduce the loss of overdue, however, there may be misjudgments when assessing the risk level of a customer, and for those high-risk customers who are misjudged as low-risk customers, a higher credit limit will be granted to them because they are determined to be high-risk customers, resulting in a higher loss of overdue, and the risk assessment effect is poor. SUMMARY

[0004] The main purpose of the present application is to provide a risk assessment method, device, electronic device and storage medium, which aims to solve the technical problem of poor risk assessment effect in the prior art.

[0005] To achieve the above-mentioned purpose, the present application provides a risk assessment method, comprising the following steps:

[0006] obtaining a sample to be evaluated;

[0007] obtaining an initial risk assessment result of the sample to be evaluated by inputting the sample to be evaluated into a first risk model;

[0008] if it is determined based on the initial risk assessment result that the sample to be evaluated belongs to a high-score customer group, then obtaining a target risk assessment result of the sample to be evaluated by inputting the sample to be evaluated into a second risk model, wherein the second risk model is trained based on target training samples belonging to the high-score customer group in the training samples of the first risk model.

[0009] The present application also provides a risk assessment device, comprising:

[0010] an acquisition module for acquiring a sample to be evaluated;

[0011] The first risk assessment module is configured to input the sample to be evaluated into a first risk model to obtain an initial risk assessment result of the sample to be evaluated.

[0012] The second risk assessment module is configured to input the sample to be evaluated into a second risk model to obtain a target risk assessment result of the sample to be evaluated if the sample to be evaluated belongs to the high-score customer group based on the initial risk assessment result, wherein the second risk model is trained based on target training samples belonging to the high-score customer group in training samples of the first risk model.

[0013] The application further provides an electronic device, which is a physical device and comprises a memory, a processor and a program of the risk assessment method stored in the memory and executable on the processor, wherein the program of the risk assessment method can realize the steps of the risk assessment method when executed by the processor.

[0014] The application further provides a storage medium, which is a computer readable storage medium and stores a program of a risk assessment method, wherein the program of the risk assessment method can realize the steps of the risk assessment method when executed by a processor.

[0015] The application further provides a computer program product comprising a computer program, wherein the computer program can realize the steps of the risk assessment method when executed by a processor.

[0016] The application provides a risk assessment method and device, electronic equipment and a storage medium. The initial risk assessment result of the to-be-evaluated sample is obtained by inputting the to-be-evaluated sample into a first risk model, so as to achieve the purpose of preliminary risk assessment of the to-be-evaluated sample based on the first risk model. Then, if it is determined based on the initial risk assessment result that the to-be-evaluated sample belongs to a high-score customer group, the target risk assessment result of the to-be-evaluated sample is obtained by inputting the to-be-evaluated sample into a second risk model. The second risk model is trained based on target training samples belonging to the high-score customer group in the training samples of the first risk model, so as to achieve the purpose of further risk assessment of the to-be-evaluated sample based on the second risk model in the case that the to-be-evaluated sample belongs to the high-score customer group. The low-score customers misjudged as belonging to the high-score customer group in the high-score customer group have certain differences from the low-score customer group, otherwise they would not be misjudged as belonging to the high-score customer group. However, the proportion of these different features is small or the weight of these different features in the low-score customer group is low, so the first risk model cannot accurately identify the low-score customers misjudged as belonging to the high-score customer group. The second risk model is trained based on the target training samples belonging to the high-score customer group in the training samples of the first risk model. Therefore, the second risk model trained based on the high-score customer group is not affected by the low-score customer group, so it can identify the low-score customers misjudged as belonging to the high-score customer group due to the small proportion of different features or the low weight of different features. Therefore, the high credit limit given to the low-score customers misjudged as belonging to the high-score customer group can effectively reduce the high overdue loss, overcome the technical defect that the risk level of customers is misjudged during risk assessment, improve the risk assessment effect, and improve the risk assessment effect. BRIEF DESCRIPTION OF DRAWINGS

[0017] The accompanying drawings, which are incorporated herein and form a part of the specification, illustrate embodiments consistent with the present application and, together with the description, further serve to explain the principles of the application.

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, brief introductions will be given to the drawings needed in the embodiments or prior art descriptions. Obviously, for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0019] Figure 1 A flowchart of an embodiment of the risk assessment method in the application;

[0020] Figure 2A schematic diagram of a first score distribution and a second score distribution of one example of the risk assessment method in the present application;

[0021] Figure 3 A flowchart of another embodiment of the risk assessment method in the present application;

[0022] Figure 4 A structural schematic diagram of one embodiment of the risk assessment device in the present application;

[0023] Figure 5 A device structural schematic diagram of a hardware running environment involved in the risk assessment method in the embodiment of the present application.

[0024] The purposes, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0025] In order to make the above objectives, features and advantages of the present application more apparent, clear and complete, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0026] Embodiment One

[0027] The embodiment of the present application provides a risk assessment method. In the first embodiment of the risk assessment method in the present application, with reference to Figure 1 , the method comprises the following steps:

[0028] Step S10, obtaining a sample to be evaluated;

[0029] The execution subject of the method in the embodiment can be a risk assessment device, a risk assessment terminal device or a server. The embodiment takes the risk assessment device as an example, which can be integrated on a terminal device with data processing function, such as a smart phone, a tablet computer and the like.

[0030] In this embodiment, it should be noted that the risk model refers to a quantitative model developed based on mathematical statistics method with the goal of identifying whether a customer will default, and is commonly used in credit risk management, such as loan approval, credit limit, risk pricing, etc. The risk model is developed from the perspective of natural persons, i.e. customers. For example, if a person repays on time, it can be marked as 0, indicating that it is a good customer. If a person has a long overdue behavior, it can be marked as 1, indicating that it is a bad customer. Further, based on the feature data of these natural persons or customers as samples, a classification model is established through classification algorithms such as logistic regression, GBDT (Gradient Boosting Decision Tree), XGBoost (eXtreme Gradient Boosting), etc.

[0031] Specifically, a to-be-evaluated sample is obtained, wherein the to-be-evaluated sample contains credit feature data related to credit of a customer to be evaluated, and the credit feature data includes one or more of a loan type, a loan amount, a overdue time, an overdue amount, and multi-loan.

[0032] In step S20, an initial risk evaluation result of the to-be-evaluated sample is obtained by inputting the to-be-evaluated sample into the first risk model.

[0033] In this embodiment, specifically, credit feature data contained in the to-be-evaluated sample is spliced into a credit feature vector. For example, assuming that the credit feature vector is (a, b, c), the feature value a represents that the loan type is a housing loan, the feature value b represents that the overdue time is 30 days, and the feature value c represents that loan demand is submitted to three financial institutions. An initial risk evaluation result of the to-be-evaluated sample is obtained by inputting the credit feature vector into a first risk model. The first risk model can be a classification model established by a classification algorithm such as logistic regression, GBDT, XGBoost, etc. The initial risk evaluation result at least includes at least one of an initial classification label, an initial classification probability, an initial quality score, etc. The initial quality score is a quality score of the to-be-evaluated sample obtained by the first risk model after data processing according to the credit feature vector. The initial classification label can include a high-score customer group, a low-score customer group, etc. The high-score customer group is composed of customers whose initial quality scores are higher than an initial quality score threshold. The low-score customer group is composed of customers whose initial quality scores are not higher than the initial quality score threshold. The initial risk evaluation result can also be an initial risk evaluation vector. For example, it is determined by the first risk model that the initial classification label of the to-be-evaluated sample is 1, indicating that the to-be-evaluated sample belongs to the high-score customer group, it is determined that the to-be-evaluated sample has an 80% probability of belonging to the high-score customer group, and the initial quality score of the to-be-evaluated sample is 900. The initial classification label 1 or the initial quality score 900 can be taken as the initial risk evaluation result. The vector (1, 0.8, 900), the vector (1, 0.8), or the vector (1, 900) can also be taken as the initial risk evaluation result.

[0034] In step S30, if it is determined based on the initial risk evaluation result that the to-be-evaluated sample belongs to the high-score customer group, a target risk evaluation result of the to-be-evaluated sample is obtained by inputting the to-be-evaluated sample into a second risk model. The second risk model is trained based on target training samples belonging to the high-score customer group in training samples of the first risk model.

[0035] In this embodiment, it should be noted that the second risk model can be a classification model established by a classification algorithm such as logistic regression, GBDT, XGBoost, etc. The second risk model is trained based on target training samples belonging to the high-score customer group in training samples of the first risk model.

[0036] Specifically, after obtaining the initial risk assessment result of the to-be-evaluated sample, it is determined whether the to-be-evaluated sample belongs to a high-score customer group according to a classification label or a quality score in the initial risk assessment result. If it is determined that the to-be-evaluated sample belongs to a high-score customer group based on the initial risk assessment result, a target risk assessment result of the to-be-evaluated sample is obtained by inputting a credit feature vector corresponding to the to-be-evaluated sample into a second risk model, so that a user determines a credit limit of the to-be-evaluated sample according to the target risk assessment result. The target risk assessment result at least includes at least one of a target classification label, a target classification probability, a target quality score, etc. The target quality score is a quality score of the to-be-evaluated sample obtained by the second risk model after data processing of the credit feature vector. The target classification label can include a high-risk high-score customer group, a low-risk high-score customer group, etc. The high-risk high-score customer group is composed of customers whose initial quality score is higher than an initial quality score threshold and whose target quality score is lower than a target quality score threshold. The low-risk high-score customer group is composed of customers whose initial quality score is higher than the initial quality score threshold and whose target quality score is not lower than the target quality score threshold. The target risk assessment result can also be a target risk assessment vector. For example, it is determined that the target classification label of the to-be-evaluated sample is 1, indicating that the to-be-evaluated sample belongs to a high-risk high-score customer group, it is determined that the to-be-evaluated sample has a probability of 75% of belonging to a high-risk high-score customer group, and the target quality score of the to-be-evaluated sample is 350. Then, the target classification label 1 or the target quality score 350 can be taken as the target risk assessment result, or the vector (1, 0.75, 350), the vector (1, 0.75), or the vector (1, 350) can be taken as the target risk assessment result. The information of the quality score in the target risk assessment result can enable the user to not only determine whether to grant credit, but also determine a specific credit amount. The higher the quality score, the higher the credit amount, and vice versa.

[0037] Optionally, the step of obtaining the target risk assessment result of the to-be-evaluated sample by inputting the to-be-evaluated sample into the second risk model comprises:

[0038] Step S31, obtaining a first quality score of the to-be-evaluated sample by inputting the to-be-evaluated sample into the second risk model.

[0039] In this embodiment, specifically, the first quality score of the to-be-evaluated sample is obtained by inputting the credit feature vector corresponding to the to-be-evaluated sample into the second risk model.

[0040] Step S32, determining a second quality score corresponding to the first quality score according to a score mapping relationship between the first risk model and the second risk model.

[0041] In the embodiment, specifically, the score mapping relationship between the first risk model and the second risk model is determined according to the quality score information of the high-score customer group in the first risk model and the quality score information of the customer group in the second risk model, the score mapping relationship between the first risk model and the second risk model is used to map the quality score of the customer group in the second risk model to the quality score range of the high-score customer group in the first risk model, and then the second quality score corresponding to the first quality score can be determined according to the preset score mapping relationship between the first risk model and the second risk model, wherein the second quality score is in the quality score range of the high-score customer group in the first risk model, and the score mapping relationship can be a score mapping relationship table, a score mapping relationship function or the like.

[0042] For example, if the score range of the first risk model is 0-100 points, the high-score customer group is composed of customers with a third quality score higher than 70 points, the quality score range of the high-score customer group in the first risk model is 70-100 points, and the quality score range of the second risk model is 0-100 points, the first quality score is between 0-100 points, and if the first quality score is directly output as a target risk assessment result, high-score customers with a score lower than 70 points may appear, which may cause the user to confuse the high-score customers with a low first quality score with other low-score customers when granting credit, and the user determines the credit amount based on the quality score, which may cause the credit amount to be incorrect. Therefore, the first quality score can be mapped to 70-100 points, so that the first quality score is comparable and clear to the third quality score of other low-score customer groups, and the user can simply and clearly know the specific situation after seeing the quality score, and is not easy to confuse and does not need to confirm in combination with other more information.

[0043] In an implementable manner, the score mapping relationship can be determined according to the proportional relationship between the quality score range of the customer group in the second risk model and the quality score range of the high-score customer group in the first risk model, for example, the quality score range of the customer group in the second risk model is 0-100 points, the quality score range of the high-score customer group in the first risk model is 80-100 points, and the second quality score can be obtained by reducing the first quality score by five times and adding 80; the score mapping relationship can also be determined according to the relationship between the quality score distribution of the customer group in the second risk model and the quality score distribution of the high-score customer group in the first risk model.

[0044] Optionally, before the step of determining the second quality score corresponding to the first quality score according to the preset score mapping relationship between the first risk model and the second risk model, the method further comprises:

[0045] Step S321, obtaining a first score distribution of the high-score customer group in the first risk model and a second score distribution of the customer group in the second risk model;

[0046] In this embodiment, specifically, after the first risk model is trained, a score distribution of the customer group in the first risk model is obtained, a first score distribution of the high-score customer group is extracted from the score distribution of the customer group in the first risk model, after the second risk model is trained, a second score distribution of the customer group in the second risk model is obtained, wherein the score distribution of the customer group in the first risk model is subject to a normal distribution, the first score distribution can be a part of a quantity decreasing area in the normal distribution, and the second score distribution is also subject to a normal distribution.

[0047] Step S322, determining a distribution mapping relationship between the first score distribution and the second score distribution as a score mapping relationship between the first risk model and the second risk model.

[0048] In this embodiment, specifically, a distribution mapping relationship is determined according to the first score distribution and the second score distribution, and the distribution mapping relationship is determined as a score mapping relationship between the first risk model and the second risk model, wherein the distribution mapping relationship is used to represent a change of the score of each customer in the case of keeping the total number of customers in the customer group unchanged and adjusting the score distribution of the customer group from the second score distribution to the first score distribution.

[0049] For example, referring to FIG. 6, Figure 2 , Figure 2 In FIG. 6, the abscissa of the two score distributions is the score, and the ordinate is the number of customers. According to the first quality score of the high-score customer group after being re-scored by the second risk model, the high-score customer group can be reordered, that is, the customer whose third quality score is originally high may have a lower first quality score, and then the second score distribution can be formed. The customers reordered according to the first quality score are arranged in order to form the first score distribution of the high-score customer group in the first risk model, and the first score distribution after reordering is obtained. After the first score distribution is formed, the second quality score of each customer can be determined according to the position of each customer in the first score distribution. In this way, the high-score customer group is reordered after being re-evaluated, which can reduce the misjudgment range of the high-score customer group, and can help the user to reduce the credit amount granted to the high-risk high-score customer with low actual quality who is misjudged as having a high score, and reduce the overdue loss.

[0050] Step S33, determining the second quality score as the target risk evaluation result of the to-be-evaluated sample.

[0051] In the embodiment, specifically, the second quality score is determined as the target risk assessment result of the sample to be evaluated, so that the user determines the credit limit of the sample to be evaluated according to the second quality score.

[0052] Optionally, after the step of obtaining the initial risk assessment result of the sample to be evaluated by inputting the sample to be evaluated into the first risk model, the method further includes:

[0053] If it is determined that the sample to be evaluated does not belong to the high-score customer group based on the initial risk assessment result, a third quality score in the initial risk assessment result is determined as the target risk assessment result of the sample to be evaluated.

[0054] In the embodiment, specifically, after obtaining the initial risk assessment result of the sample to be evaluated, it is determined whether the sample to be evaluated belongs to the high-score customer group according to a classification label or a quality score in the initial risk assessment result. If it is determined that the sample to be evaluated does not belong to the high-score customer group based on the initial risk assessment result, a third quality score in the initial risk assessment result is determined as the target risk assessment result of the sample to be evaluated. The case that the sample to be evaluated does not belong to the high-score customer group can be that the sample to be evaluated belongs to a low-score customer group, a pass-score customer group, or the like. The specific case can be determined according to actual needs and the number of classification categories of the first risk model, which is not limited in the embodiment. Determining the quality score as the final target risk assessment result can enable the user to not only determine whether to grant credit but also determine the specific credit amount. The higher the quality score is, the higher the credit amount is, and vice versa.

[0055] Optionally, the initial risk assessment result includes a third quality score. After the step of obtaining the initial risk assessment result of the sample to be evaluated by inputting the sample to be evaluated into the first risk model, the method further includes:

[0056] Step A10, comparing the third quality score with a preset quality score threshold;

[0057] Step A20, if it is determined that the third quality score is greater than the preset quality score threshold, it is determined that the sample to be evaluated belongs to the high-score customer group.

[0058] Step A30, if it is determined that the third quality score is less than or equal to the preset quality score threshold, it is determined that the sample to be evaluated belongs to the low-score customer group.

[0059] In this embodiment, specifically, after obtaining the initial risk assessment result of the to-be-evaluated sample, a third quality score of the to-be-evaluated sample is extracted from the initial risk assessment result, and a size relationship between the third quality score and a preset quality score threshold is compared, wherein the preset quality score threshold can be set according to actual conditions, can be determined according to the actual proportion of high-score customers, or can be determined according to business experience, and the present embodiment does not limit this. Further, if it is determined that the third quality score is greater than the preset quality score threshold, it is determined that the to-be-evaluated sample belongs to the high-score customer group, and if it is determined that the third quality score is less than or equal to the preset quality score threshold, it is determined that the to-be-evaluated sample belongs to the low-score customer group.

[0060] In this embodiment, by obtaining the to-be-evaluated sample, by inputting the to-be-evaluated sample into the first risk model, the initial risk assessment result of the to-be-evaluated sample is obtained, the purpose of preliminarily evaluating the risk of the to-be-evaluated sample based on the first risk model is achieved, and further, if it is determined based on the initial risk assessment result that the to-be-evaluated sample belongs to the high-score customer group, the to-be-evaluated sample is input into the second risk model to obtain the target risk assessment result of the to-be-evaluated sample, wherein the second risk model is trained based on the target training sample belonging to the high-score customer group in the training sample of the first risk model, the purpose of further evaluating the risk of the to-be-evaluated sample based on the second risk model in the case that the to-be-evaluated sample belongs to the high-score customer group is achieved. The low-score customers misjudged as belonging to the high-score customer group in the high-score customer group have certain differences from the low-score customer group, otherwise they would not be misjudged as belonging to the high-score customer group, but the proportion of these different features is small or the weight in the low-score customer group is low, so in the first risk model, the low-score customers misjudged as belonging to the high-score customer group cannot be accurately identified. The second risk model is trained based on the target training sample belonging to the high-score customer group in the training sample of the first risk model, so the second risk model trained only based on the high-score customer group will not be affected by the low-score customer group, and thus the low-score customers misjudged as belonging to the high-score customer group due to the small proportion of different features or low weight can be identified from the high-score customer group. Further, the high risk customers misjudged as low risk customers will be given a high credit limit due to being determined as high risk customers, resulting in high overdue loss and poor risk assessment effect, and the technical defect of poor risk assessment effect is overcome, and the risk assessment effect is improved.

[0061] Embodiment Two

[0062] Further, with reference to Figure 3, based on the above embodiments of the application, in the second embodiment of the application, the same or similar as the above embodiments, can be referred to the above introduction, the following will not be described. On this basis, the step of inputting the to-be-evaluated sample into the first risk model to obtain the initial risk evaluation result of the to-be-evaluated sample, further comprises:

[0063] Step B10, obtaining training samples and training sample labels corresponding to the training samples;

[0064] In this embodiment, specifically, the training samples are obtained, and the training samples are artificially labeled to obtain the training sample labels corresponding to the training samples. The training sample labels at least include at least one of training sample classification labels, training sample classification probabilities, training sample quality scores, etc.

[0065] Step B20, by inputting the training sample into the first risk model to be trained, determining the first training model prediction result of the training sample;

[0066] In this embodiment, specifically, the credit feature training data contained in the training sample is spliced into a credit feature training vector, and the credit feature training vector is input into the first risk model to be trained to determine the first training model prediction result of the training sample. The first training model prediction result at least includes at least one of initial training classification labels, initial training classification probabilities, initial training quality scores, etc. The initial training quality score is the quality score of the training sample obtained by the first risk model after data processing according to the credit feature training vector. The initial classification label can include high-score customer group, low-score customer group, etc. The high-score customer group is composed of customers whose training quality score is higher than the training quality score threshold. The low-score customer group is composed of customers whose training quality score is not higher than the training quality score threshold. The training risk evaluation result can also be a training risk evaluation vector.

[0067] Step B30, based on the training sample labels and the first training model prediction result, by calculating the first model prediction loss corresponding to the first risk model to be trained, iteratively optimizing the first risk model to be trained to obtain the first risk model;

[0068] In the embodiment, specifically, the first model prediction loss corresponding to the first risk model to be trained is calculated according to the difference between the training sample label and the first training model prediction result; it is judged whether the first model prediction loss converges, if the first model prediction loss converges, the first risk model to be trained is taken as the first risk model; if the first model prediction loss does not converge, the first risk model to be trained is updated according to the model gradient calculated by the first model prediction loss, and the step of obtaining the training sample and the training sample label corresponding to the training sample is executed until the first model prediction loss calculated converges.

[0069] Optionally, before the step of obtaining the target risk evaluation result of the sample to be evaluated by inputting the sample to be evaluated into the second risk model, the method further comprises:

[0070] Step C10, determining a target training sample belonging to a high-score customer group from the training samples;

[0071] In the embodiment, specifically, after the training of the first risk model is completed, the target training sample belonging to the high-score customer group is extracted from the training samples according to the evaluation result of the first risk model.

[0072] Step C20, inputting the target training sample into a second risk model to be trained to determine a second training model prediction result of the target training sample;

[0073] In the embodiment, specifically, the target training sample corresponding to the target credit feature training vector spliced by the first risk model during training can be directly obtained, or the credit feature training data contained in the target training sample can be spliced into a target credit feature training vector, and then the target credit feature training vector is input into the second risk model to be trained to determine the second training model prediction result of the target training sample. The second risk model to be trained can be the same as the first risk model to be trained, that is, the same model is trained by different training samples. The second training model prediction result includes at least one of a target training classification label, a target training classification probability, a target training quality score, etc. The target training quality score is the target training quality score of the target training sample obtained by the second risk model after data processing of the target credit feature training vector. The target initial classification label can include a high-risk high-score customer group and a low-risk high-score customer group. The high-risk high-score customer group is composed of customers whose training quality score is higher than a training quality score threshold and whose target training quality score is lower than a target training quality score threshold. The low-risk high-score customer group is composed of customers whose training quality score is higher than the training quality score threshold and whose target training quality score is not lower than the target training quality score threshold. The target training risk assessment result can also be a target training risk assessment vector.

[0074] In step C30, based on the training sample label corresponding to the target training sample and the second training model prediction result, the second model prediction loss corresponding to the second risk model to be trained is calculated, and the second risk model to be trained is iteratively optimized to obtain a second risk model.

[0075] In the embodiment, specifically, the second model prediction loss corresponding to the second risk model to be trained is calculated according to the difference between the training sample label corresponding to the target training sample and the second training model prediction result. It is judged whether the second model prediction loss converges. If the second model prediction loss converges, the second risk model to be trained is taken as a second risk model. If the second model prediction loss does not converge, the second risk model to be trained is updated according to the model gradient calculated by the second model prediction loss, and the step of determining the target training sample belonging to the high-score customer group from the training sample is performed until the second model prediction loss converges.

[0076] In the embodiment, by taking the high-score customer group in the first risk model as the training sample of the second risk model, the trained second risk model is not affected by the low-score customer group, and thus the low-score customers originally misjudged as belonging to the high-score customer group due to the small proportion of the difference feature or the low weight can be identified from the high-score customer group, thereby effectively reducing the high overdue loss possibly caused by giving the low-score customers misjudged as belonging to the high-score customer group a high credit limit, overcoming the technical defect that the risk level of the customer is misjudged during the evaluation, and improving the risk evaluation effect.

[0077] Embodiment three

[0078] Further, the embodiment of the application further provides a risk evaluation device, referring to Figure 4 , the risk evaluation device is applied to a risk evaluation party, and comprises:

[0079] The first acquisition module 10 is configured to acquire a to-be-evaluated sample.

[0080] The first risk evaluation module 20 is configured to input the to-be-evaluated sample into a first risk model to obtain an initial risk evaluation result of the to-be-evaluated sample.

[0081] The second risk evaluation module 30 is configured to, if it is determined that the to-be-evaluated sample belongs to a high-score customer group based on the initial risk evaluation result, input the to-be-evaluated sample into a second risk model to obtain a target risk evaluation result of the to-be-evaluated sample, wherein the second risk model is trained based on target training samples belonging to the high-score customer group in training samples of the first risk model.

[0082] Optionally, the first risk evaluation module 20 is further configured to:

[0083] input the to-be-evaluated sample into the second risk model to obtain a first quality score of the to-be-evaluated sample;

[0084] determine a second quality score corresponding to the first quality score according to a preset score mapping relationship between the first risk model and the second risk model;

[0085] determine the second quality score as the target risk evaluation result of the to-be-evaluated sample.

[0086] Optionally, before the step of determining the second quality score corresponding to the first quality score according to the preset score mapping relationship between the first risk model and the second risk model, the first risk evaluation module 20 is further configured to:

[0087] obtain a first score distribution of the high-score customer group in the first risk model and a second score distribution of the customer group in the second risk model;

[0088] determine a distribution mapping relationship between the first score distribution and the second score distribution as a score mapping relationship between the first risk model and the second risk model.

[0089] Optionally, the risk assessment device further comprises a third risk assessment module, after the step of obtaining the initial risk assessment result of the to-be-evaluated sample by inputting the to-be-evaluated sample into the first risk model, the third risk assessment module is configured to:

[0090] if it is determined based on the initial risk assessment result that the to-be-evaluated sample does not belong to the high-score customer group, determining a third quality score in the initial risk assessment result as the target risk assessment result of the to-be-evaluated sample.

[0091] Optionally, the risk assessment device further comprises a training module, before the step of obtaining the initial risk assessment result of the to-be-evaluated sample by inputting the to-be-evaluated sample into the first risk model, the training module is configured to:

[0092] obtain a training sample and a training sample label corresponding to the training sample;

[0093] determine a first training model prediction result of the training sample by inputting the training sample into the to-be-trained first risk model;

[0094] based on the training sample label and the first training model prediction result, iteratively optimize the to-be-trained first risk model by calculating a first model prediction loss corresponding to the to-be-trained first risk model, to obtain the first risk model.

[0095] Optionally, before the step of obtaining the target risk assessment result of the to-be-evaluated sample by inputting the to-be-evaluated sample into the second risk model, the training module is further configured to:

[0096] determine a target training sample belonging to the high-score customer group from the training sample;

[0097] input the target training sample into the to-be-trained second risk model to determine a second training model prediction result of the target training sample;

[0098] based on the training sample label corresponding to the target training sample and the second training model prediction result, iteratively optimize the to-be-trained second risk model by calculating a second model prediction loss corresponding to the to-be-trained second risk model, to obtain the second risk model.

[0099] Optionally, the risk assessment device further comprises a determination module, when the initial risk assessment result comprises a third quality score, after the step of obtaining the initial risk assessment result of the sample to be assessed by inputting the sample to be assessed into the first risk model, the determination module is used for:

[0100] comparing the third quality score with a preset quality score threshold;

[0101] if it is determined that the third quality score is greater than the preset quality score threshold, determining that the sample to be assessed belongs to a high-score customer group;

[0102] if it is determined that the third quality score is less than or equal to the preset quality score threshold, determining that the sample to be assessed belongs to a low-score customer group.

[0103] The risk assessment device provided by the present application adopts the risk assessment method in the above embodiments, and solves the technical problem of poor risk assessment effect in the prior art. Compared with the prior art, the risk assessment device provided by the embodiments of the present application has the same beneficial effects as the risk assessment method provided by the above embodiments, and other technical features in the risk assessment device are the same as the features disclosed in the above embodiments, which will not be repeated here.

[0104] Embodiment Four

[0105] Further, the embodiments of the present application provide an electronic device, which comprises at least one processor and a memory in communication connection with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the risk assessment method or the conversion qualification cutoff parameter determination method in the above embodiments.

[0106] Reference will now be made to Figure 5 which shows a structural schematic diagram of an electronic device suitable for implementing the embodiments of the present disclosure. The electronic device in the embodiments of the present disclosure can include but is not limited to mobile terminals such as Bluetooth headsets, mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablets), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), and the like, and fixed terminals such as digital TVs, desktop computers, and the like. Figure 5 The electronic device shown is only an example, and should not impose any limitation on the functions and use range of the embodiments of the present disclosure.

[0107] As Figure 5As shown, the electronic device can include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) that can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) or a program loaded into a random access memory (RAM) from a storage device. In the RAM, various programs and arrays required for the operation of the electronic device are also stored. The processing device, the ROM, and the RAM are connected to each other through a bus. An input / output (I / O) interface is also connected to the bus.

[0108] In general, the following systems can be connected to the I / O interface: input devices including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices including, for example, a magnetic tape, a hard disk, etc.; and communication devices. The communication devices can allow the electronic device to communicate wirelessly or wiredly with other devices to exchange arrays. While the electronic device having various systems is shown in the drawing, it is understood that all of the shown systems are not required to be implemented or possessed. More or less systems can be alternatively implemented or possessed.

[0109] In particular, according to embodiments of the present disclosure, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, embodiments of the present disclosure include a computer program product including a computer program carried on a computer readable medium, the computer program containing program codes for executing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network through the communication devices, or installed from the storage devices, or installed from the ROM. When the computer program is executed by the processing device, the above-mentioned functions defined in the methods of the embodiments of the present disclosure are performed.

[0110] The electronic device provided by the present application adopts the risk assessment method or the conversion qualification cutoff parameter determination method in the above embodiments, which solves the technical problem of poor risk assessment effect in the prior art. Compared with the prior art, the electronic device provided by the embodiments of the present application has the same beneficial effects as the risk assessment method or the conversion qualification cutoff parameter determination method provided by the above embodiments, and other technical features in the electronic device are the same as the features disclosed in the above method embodiments, which will not be repeated here.

[0111] It should be understood that parts of the present disclosure can be realized by hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0112] The above merely illustrates the specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which shall be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

[0113] Embodiment six

[0114] Further, the embodiment provides a computer readable storage medium having stored thereon computer readable program instructions for performing the risk assessment method or the conversion qualification cut-off parameter determination method in the above embodiments.

[0115] The computer readable storage medium provided by the embodiment of the present application may, for example, be a U disk, but is not limited to an electric, magnetic, optical, electromagnetic, infrared, or semiconductor system, system, or device, or any combination thereof. More specific examples of the computer readable storage medium can include, but are not limited to, an electric connection with one or more conductive wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the embodiment, the computer readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer readable storage medium can be transmitted by any suitable medium, including but not limited to an electric wire, an optical cable, an RF (radio frequency), etc., or any suitable combination thereof.

[0116] The above computer readable storage medium can be contained in an electronic device; or can exist separately without being assembled into an electronic device.

[0117] The above computer readable storage medium carries one or more programs, which, when executed by an electronic device, cause the electronic device to: obtain a sample to be evaluated; obtain an initial risk assessment result of the sample to be evaluated by inputting the sample to be evaluated into a first risk model; and if it is determined based on the initial risk assessment result that the sample to be evaluated belongs to a high-score customer group, obtain a target risk assessment result of the sample to be evaluated by inputting the sample to be evaluated into a second risk model, wherein the second risk model is trained based on target training samples belonging to the high-score customer group in training samples of the first risk model.

[0118] Computer program code for carrying out operations of the present disclosure can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0119] The computer program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0120] The modules involved in the embodiments of the present disclosure can be implemented in the manner of software or hardware. In some cases, the name of the module does not constitute a limitation on the module itself.

[0121] The computer readable storage medium provided by the present application stores computer readable program instructions for executing the risk assessment method or the conversion qualification cutoff parameter determination method, and solves the technical problem of poor risk assessment effect in the prior art. Compared with the prior art, the computer readable storage medium provided by the embodiment of the present application has the same beneficial effects as the risk assessment method or the conversion qualification cutoff parameter determination method provided by the above-mentioned embodiment, and will not be described here.

[0122] Embodiment seven

[0123] Further, the present application also provides a computer program product comprising a computer program which, when executed by a processor, implements the steps of the risk assessment method or the conversion qualification cut-off parameter determination method as described above.

[0124] The computer program product provided by the present application solves the technical problem of poor risk assessment effect in the prior art. Compared with the prior art, the computer program product provided by the embodiments of the present application has the same beneficial effects as the risk assessment method or the conversion qualification cut-off parameter determination method provided by the above-mentioned embodiments, and will not be described here.

[0125] The above is only the preferred embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation using the content of the present application specification and drawings, or direct or indirect application in other related technical fields, is also included in the patent processing scope of the present application.

Claims

1. A risk assessment method, characterized by, The risk assessment method comprises the following steps: obtaining a to-be-evaluated sample; obtaining an initial risk assessment result of the to-be-evaluated sample by inputting the to-be-evaluated sample into a first risk model; if it is determined based on the initial risk assessment result that the to-be-evaluated sample belongs to a high-score customer group, obtaining a target risk assessment result of the to-be-evaluated sample by inputting the to-be-evaluated sample into a second risk model, wherein the second risk model is trained based on target training samples belonging to the high-score customer group in training samples of the first risk model; the step of obtaining the target risk assessment result of the to-be-evaluated sample by inputting the to-be-evaluated sample into the second risk model comprises: obtaining a first quality score of the to-be-evaluated sample by inputting the to-be-evaluated sample into the second risk model; obtaining a first score distribution of the high-score customer group in the first risk model and a second score distribution of the customer group in the second risk model; determining a distribution mapping relationship between the first score distribution and the second score distribution as a score mapping relationship between the first risk model and the second risk model; determining a second quality score corresponding to the first quality score according to a pre-set score mapping relationship between the first risk model and the second risk model; wherein the high-score customer group is reordered according to the first quality score after the high-score customer group is re-scored by the second risk model, the customers reordered according to the first quality score are arranged in order into the first score distribution of the high-score customer group in the first risk model, a first score distribution with the same score but reordered is obtained, and the second quality score corresponding to each customer is determined according to the position of the customer in the first score distribution after the first score distribution is formed; determining the second quality score as the target risk assessment result of the to-be-evaluated sample.

2. The risk assessment method of claim 1, wherein, after the step of obtaining the initial risk assessment result of the to-be-evaluated sample by inputting the to-be-evaluated sample into the first risk model, the method further comprises: if it is determined based on the initial risk assessment result that the to-be-evaluated sample does not belong to the high-score customer group, determining a third quality score in the initial risk assessment result as the target risk assessment result of the to-be-evaluated sample.

3. The risk assessment method of claim 1, wherein, before the step of obtaining the initial risk assessment result of the to-be-evaluated sample by inputting the to-be-evaluated sample into the first risk model, the method further comprises: obtaining training samples and training sample labels corresponding to the training samples; determining a first training model prediction result of the training samples by inputting the training samples into a first risk model to be trained; based on the training sample labels and the first training model prediction result, the first risk model is obtained by iteratively optimizing the first risk model corresponding to the first risk model prediction loss.

4. The risk assessment method of claim 3, wherein, before the step of obtaining the target risk assessment result of the to-be-evaluated sample by inputting the to-be-evaluated sample into the second risk model, the method further comprises: determining target training samples belonging to the high-score customer group from the training samples; inputting the target training sample into a second risk model to be trained to determine a second training model prediction result of the target training sample; based on the training sample label corresponding to the target training sample and the second training model prediction result, a second model prediction loss corresponding to the second risk model to be trained is calculated, and the second risk model to be trained is iteratively optimized to obtain a second risk model.

5. The risk assessment method of claim 1, wherein, The initial risk assessment result includes a third quality score, and after the step of inputting the sample to be evaluated into the first risk model to obtain the initial risk assessment result of the sample to be evaluated, the method further comprises: comparing the third quality score with a preset quality score threshold; if it is determined that the third quality score is greater than the preset quality score threshold, it is determined that the sample to be evaluated belongs to a high-score customer group; if it is determined that the third quality score is less than or equal to the preset quality score threshold, it is determined that the sample to be evaluated belongs to a low-score customer group.

6. A risk assessment device, comprising: an acquisition module configured to acquire a sample to be evaluated; a first risk assessment module configured to input the sample to be evaluated into a first risk model to obtain an initial risk assessment result of the sample to be evaluated; a second risk assessment module configured to, if it is determined based on the initial risk assessment result that the sample to be evaluated belongs to a high-score customer group, input the sample to be evaluated into a second risk model to obtain a target risk assessment result of the sample to be evaluated, wherein the second risk model is trained based on target training samples belonging to a high-score customer group in training samples of the first risk model; and further configured to: input the sample to be evaluated into the second risk model to obtain a first quality score of the sample to be evaluated; acquire a first score distribution of the high-score customer group in the first risk model and a second score distribution of the customer group in the second risk model; determine a distribution mapping relationship between the first score distribution and the second score distribution as a score mapping relationship between the first risk model and the second risk model; determine a second quality score corresponding to the first quality score according to a preset score mapping relationship between the first risk model and the second risk model; wherein the high-score customer group is reordered according to the first quality score of the high-score customer group after being re-scored by the second risk model, the customers reordered according to the first quality score are sequentially arranged into the first score distribution of the high-score customer group in the first risk model to obtain the first score distribution after reordering with the same score, and the second quality score corresponding to each customer is determined according to the position of the customer in the first score distribution after the first score distribution is formed; and determine the second quality score as the target risk assessment result of the sample to be evaluated.

7. An electronic device, comprising: The electronic device comprises: at least one processor; and a memory in communication connection with the at least one processor; wherein The memory stores instructions executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of the risk assessment method of any one of claims 1 to 5.

8. A storage medium, characterized by The storage medium is a computer readable storage medium, and the computer readable storage medium stores a program for implementing a risk assessment method, and the program for implementing a risk assessment method is executed by a processor to implement the steps of the risk assessment method of any one of claims 1 to 5.

Images