Target object risk level prediction method, device and computer equipment

Abstract

The present application relates to a method, apparatus, computer equipment, storage medium, and computer program product for predicting the risk level of a target object, which can be used in the financial field, artificial intelligence field, or other fields. The method comprises: obtaining sample data of a business scenario, training an initial first machine learning model according to a training set, obtaining a first machine learning model of the business scenario, testing the first machine learning model of the business scenario according to a test set, obtaining a prediction result, grouping the training set according to the prediction result, obtaining a grouped training set, training multiple initial second machine learning models respectively according to the grouped training set, obtaining multiple second machine learning models of the business scenario, aggregating the multiple second machine learning models of the business scenario, obtaining an interpretable machine learning model group of the business scenario, and supporting the obtaining of good interpretability results based on the interpretable machine learning model group of the business scenario.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, computer equipment, storage medium, and computer program product for predicting the risk level of a target object. Background Art

[0002] With the development of artificial intelligence technology, machine learning models have been increasingly widely used. The introduction of machine learning technology has changed the traditional model methods such as expert rules, business experience and statistical models in marketing, risk control and other fields, greatly improving the model accuracy. However, with the increasing use of complex machine learning algorithms represented by integrated algorithms and neural network algorithms, the one-sided pursuit of high precision of machine learning models cannot meet actual needs, and the interpretability of the model is also crucial.

[0003] At present, the research on the interpretability of machine learning model algorithms mainly adopts the method of building equivalent models or performing local interpretation. However, this method of building equivalent models or performing local interpretation has the problem of poor interpretability of machine learning model algorithms. Summary of the Invention

[0004] Based on this, it is necessary to address the technical problem of poor interpretability of traditional machine learning models and provide a target object risk level prediction method, device, computer equipment, computer-readable storage medium and computer program product that can support obtaining good interpretable results.

[0005] In a first aspect, the present application provides a method for predicting the risk level of a target object. The method comprises:

[0006] Obtain sample data for business scenarios, including training sets and test sets;

[0007] Training the initial first machine learning model according to the training set to obtain a first machine learning model for the business scenario, where the initial first machine learning model is a precision-oriented machine learning model;

[0008] Test the first machine learning model of the business scenario based on the test set to obtain prediction results;

[0009] According to the prediction results, the training set is grouped to obtain the grouped training set;

[0010] Training multiple initial second machine learning models according to the grouped training sets to obtain multiple second machine learning models for business scenarios, where the initial second machine learning models are interpretability-oriented machine learning models;

[0011] Aggregate multiple second machine learning models of business scenarios to obtain an explainable machine learning model group of the business scenarios.

[0012] In one embodiment, the training set is grouped according to the prediction results, and the grouped training set includes:

[0013] Sort the prediction results and combine them with the preset inter-group distance to obtain the pre-grouping of the training set;

[0014] According to the pre-grouping, the number of sample data of each pre-group is obtained;

[0015] According to the pre-grouping and the number of sample data in each pre-grouping, the grouped training set is obtained by centralized grouping.

[0016] In one embodiment, according to the pre-grouping and the number of sample data in each pre-grouping, the grouped training set is obtained by centralized grouping, including:

[0017] If the number of pre-grouped sample data is less than the number of adjacent pre-grouped samples in a preset ratio, the pre-grouped sample data is merged with the corresponding adjacent pre-grouped sample data;

[0018] If the number of pre-grouped sample data is greater than or equal to the number of adjacent pre-grouped samples in a preset ratio, a grouped training set is obtained based on the pre-grouped sample data.

[0019] In one embodiment, a plurality of second machine learning models of a business scenario are aggregated to obtain an interpretable machine learning model group of the business scenario, and then the following steps are further included:

[0020] Explain the interpretable machine learning model group of the business scenario and obtain the interpretation results of the interpretable machine learning model group of the business scenario.

[0021] In one embodiment, the target object risk level prediction method further includes:

[0022] Obtain sample data for the business scenario to be tested;

[0023] Predicting sample data of the business scenario to be tested using a first machine learning model of the business scenario to obtain a first prediction result of the business scenario to be tested;

[0024] According to the first prediction result of the business scenario to be tested and the explainable machine learning model group, a second prediction result of the business scenario to be tested is obtained.

[0025] In one embodiment, obtaining a second prediction result of the business scenario to be tested based on the first prediction result of the business scenario to be tested and the interpretable machine learning model group includes:

[0026] According to the grouped training set, the grouping space is obtained;

[0027] Determine the grouping space corresponding to the first prediction result to obtain a target grouping space;

[0028] Selecting a second machine learning model corresponding to the target grouping space based on the interpretable machine learning model group;

[0029] The sample data of the business scenario to be tested is predicted through the selected second machine learning model to obtain a second prediction result of the business scenario to be tested.

[0030] In a second aspect, the present application also provides a device for predicting the risk level of a target object. The device comprises:

[0031] The data acquisition module is used to obtain sample data for business scenarios. The sample data includes training sets and test sets.

[0032] A first machine learning model acquisition module is used to train an initial first machine learning model according to a training set to obtain a first machine learning model for a business scenario, where the initial first machine learning model is a precision-oriented machine learning model;

[0033] A prediction result acquisition module is used to test the first machine learning model of the business scenario based on the test set to obtain a prediction result;

[0034] The training set grouping module is used to group the training set according to the prediction results to obtain the grouped training set;

[0035] A second machine learning model acquisition module is used to train multiple initial second machine learning models according to the grouped training sets to obtain multiple second machine learning models for business scenarios, where the initial second machine learning models are interpretability-oriented machine learning models;

[0036] The second machine learning model collection module is used to collect multiple second machine learning models of business scenarios to obtain an explainable machine learning model group of the business scenarios.

[0037] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are performed:

[0038] Obtain sample data of the business scenario, where the sample data includes a training set and a test set; train an initial first machine learning model based on the training set to obtain a first machine learning model for the business scenario, where the initial first machine learning model is a precision-oriented machine learning model; test the first machine learning model for the business scenario based on the test set to obtain a prediction result; group the training set based on the prediction result to obtain a grouped training set; train multiple initial second machine learning models based on the grouped training set to obtain multiple second machine learning models for the business scenario, where the initial second machine learning model is an interpretability-oriented machine learning model; aggregate multiple second machine learning models for the business scenario to obtain an interpretable machine learning model group for the business scenario.

[0039] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the following steps:

[0040] Obtain sample data of the business scenario, where the sample data includes a training set and a test set; train an initial first machine learning model based on the training set to obtain a first machine learning model for the business scenario, where the initial first machine learning model is a precision-oriented machine learning model; test the first machine learning model for the business scenario based on the test set to obtain a prediction result; group the training set based on the prediction result to obtain a grouped training set; train multiple initial second machine learning models based on the grouped training set to obtain multiple second machine learning models for the business scenario, where the initial second machine learning model is an interpretability-oriented machine learning model; aggregate multiple second machine learning models for the business scenario to obtain an interpretable machine learning model group for the business scenario.

[0041] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the following steps:

[0042] Obtain sample data of the business scenario, where the sample data includes a training set and a test set; train an initial first machine learning model based on the training set to obtain a first machine learning model for the business scenario, where the initial first machine learning model is a precision-oriented machine learning model; test the first machine learning model for the business scenario based on the test set to obtain a prediction result; group the training set based on the prediction result to obtain a grouped training set; train multiple initial second machine learning models based on the grouped training set to obtain multiple second machine learning models for the business scenario, where the initial second machine learning model is an interpretability-oriented machine learning model; aggregate multiple second machine learning models for the business scenario to obtain an interpretable machine learning model group for the business scenario.

[0043] The above-mentioned target object risk level prediction method, device, computer equipment, storage medium and computer program product comprehensively use an initial first machine learning model oriented towards accuracy and an initial second machine learning model oriented towards interpretability. Specifically, the initial first machine learning model is trained with a training set, and based on the prediction results of the training set, the grouped training set is obtained by centralized grouping; the grouped training set is then used to train multiple initial second machine learning models, thereby obtaining multiple second machine learning models with excellent interpretability, and finally constructing an interpretable machine learning model group for business scenarios, which can support obtaining good interpretability results. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A diagram illustrating an application environment of a target object risk level prediction method in one embodiment;

[0045] Figure 2 Schematic diagram of a process for predicting the risk level of a target object in one embodiment;

[0046] Figure 3 Schematic diagram of a process for predicting the risk level of a target object in one embodiment;

[0047] Figure 4 is a flow chart of a method for predicting the risk level of a target object in another embodiment;

[0048] Figure 5 A schematic flow chart of a target object risk level prediction step in another embodiment;

[0049] Figure 6 Schematic diagram of a flow chart of a target object risk level prediction method in another embodiment;

[0050] Figure 7 is a structural block diagram of a target object risk level prediction device in one embodiment;

[0051] Figure 8 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0053] The target object risk level prediction method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. The terminal 102 obtains sample data of the business scenario, and the sample data includes a training set and a test set; the initial first machine learning model is trained according to the training set to obtain the first machine learning model of the business scenario, and the initial first machine learning model is a precision-oriented machine learning model; the first machine learning model of the business scenario is tested according to the test set to obtain a prediction result; according to the prediction result, the training set is grouped to obtain a grouped training set; according to the grouped training set, multiple initial second machine learning models are trained respectively to obtain multiple second machine learning models of the business scenario, and the initial second machine learning model is an explainability-oriented machine learning model; the multiple second machine learning models of the business scenario are aggregated to obtain an explainable machine learning model group of the business scenario, and the server 104 receives the explainable machine learning model group of the business scenario sent by the terminal 102. Terminal 102 may include, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices may include smart speakers, smart TVs, smart air conditioners, and smart car devices. Portable wearable devices may include smart watches, smart bracelets, and head-mounted devices. Server 104 may be implemented as a standalone server or a server cluster consisting of multiple servers.

[0054] In one embodiment, Figure 2 As shown, a method for predicting the risk level of a target object is provided. Figure 1 Taking the terminal 102 in FIG. 1 as an example, the method includes the following steps:

[0055] S100: Obtain sample data of a business scenario, where the sample data includes a training set and a test set.

[0056] Among them, the sample data of a business scenario refers to an instance of data in a specific business scenario. According to whether it has a label, the sample data can be divided into labeled samples and unlabeled samples. Labeled samples contain both features and labels, which are used to train the model. Unlabeled samples contain features but no labels. After using the labeled samples to train the model, the model is used to predict the labels of the unlabeled samples; according to the source of the data, the sample data can be roughly divided into internal samples and external samples. The methods for obtaining internal samples usually include data burial and storage, data labeling, data enhancement and data reorganization. The methods for obtaining external samples usually include public data sets, crawling and crowdsourcing, and procurement and business cooperation. This application does not specifically limit the specific sample data acquisition methods. In the field of artificial intelligence, a common practice in machine learning is to randomly divide samples into training sets and test sets from the same sample set in a certain proportion. The training set is used to train supervised models, fit models, adjust parameters, select input variables, and make other decisions about the algorithm. The test set is used to evaluate the effectiveness of the trained model, but it does not change the parameters and effects of the model. It generally verifies whether the model is overfitting or underfitting, and decides whether to retrain the model or choose another algorithm. Therefore, the training set and test set must be reasonably selected so that their distribution can represent the actual data distribution in order to expect the algorithm to have good results.

[0057] Specifically, sample data of the business scenario is obtained, and the sample data includes a training set and a test set. In this application, both the training set and the test set are labeled samples.

[0058] S200: Train the initial first machine learning model according to the training set to obtain a first machine learning model for the business scenario.

[0059] Among them, machine learning model training refers to inputting sample data into the selected machine learning model to allow the model to learn the relationship between features and labels. The selected initial first machine learning model is a precision-oriented machine learning model. The initial first machine learning model algorithm used in this application can be but is not limited to the LightGBM (Light GradientBoosting Machine) algorithm, which supports highly efficient parallel training.

[0060] Specifically, the initial first machine learning model is trained according to the training set to obtain a first machine learning model for the business scenario.

[0061] S300: Test the first machine learning model of the business scenario according to the test set to obtain a prediction result.

[0062] Among them, the trained machine learning model is applied to the sample data, that is, the trained model is used to make useful predictions on the sample data, and the prediction results of the sample data are obtained. The machine learning model is tested according to the test set to evaluate the prediction performance of the machine learning model.

[0063] Specifically, the first machine learning model of the business scenario is tested according to the test set to obtain a prediction result.

[0064] S400: Grouping the training set according to the prediction result to obtain a grouped training set.

[0065] Among them, the prediction results of the trained model for the sample data are probability data, that is, the prediction result is a real number between 0 and 1. Each training sample has a corresponding prediction result. Therefore, according to the prediction result, the training samples corresponding to the prediction result can be grouped to obtain the grouped training samples.

[0066] Specifically, the training set is grouped according to the prediction results to obtain grouped training sets.

[0067] S500: Train multiple initial second machine learning models according to the grouped training sets to obtain multiple second machine learning models for business scenarios.

[0068] Among them, the selected initial second machine learning model is an interpretability-oriented machine learning model. The initial second machine learning model used in this application may include but is not limited to linear regression or decision tree. After the training set is grouped, multiple grouped training sets are obtained. Each training set trains an initial second machine learning model, and multiple grouped training sets respectively train multiple initial second machine learning models to obtain multiple second machine learning models.

[0069] Specifically, multiple initial second machine learning models are trained separately according to the grouped training sets to obtain multiple second machine learning models for business scenarios.

[0070] S600: Aggregate multiple second machine learning models of the business scenario to obtain an explainable machine learning model group of the business scenario.

[0071] Among them, the multiple second machine learning models of the obtained business scenarios are grouped together to obtain a second machine learning model group of the business scenarios, and the second machine learning model group is the explainable machine learning model group of the business scenarios.

[0072] Specifically, multiple second machine learning models of business scenarios are aggregated to obtain an explainable machine learning model group of the business scenarios.

[0073] In the above-mentioned target object risk level prediction method, an initial first machine learning model oriented towards accuracy and an initial second machine learning model oriented towards interpretability are used in combination. Specifically, the initial first machine learning model is trained with a training set, and based on the prediction results of the training set, centralized grouping is performed to obtain a grouped training set; then, multiple initial second machine learning models are trained with the grouped training set, thereby obtaining multiple second machine learning models with excellent interpretability. Finally, an interpretable machine learning model group for business scenarios is constructed, which can support obtaining good interpretability results.

[0074] In one embodiment, Figure 3 As shown, according to the prediction results, the training set is grouped, and the grouped training sets include:

[0075] S420: Sort the prediction results and obtain the pre-grouping of the training set based on the preset inter-group distance;

[0076] S440: According to the pre-grouping, obtain the number of sample data in each pre-group;

[0077] S460: According to the pre-grouping and the number of sample data in each pre-grouping, a grouped training set is obtained by centralized grouping.

[0078] In this embodiment, the prediction results are sorted, and combined with a preset inter-group distance to obtain pre-groupings of the training set. Based on the pre-groupings, the number of sample data in each pre-group is obtained. Specifically, the prediction results are sorted according to a certain rule to obtain sorted prediction results. The sorting rule can be from small to large, and the specific sorting rule is not limited in this application. The sorted prediction results are divided into n groups according to the preset component distance, and the training set is also grouped into n groups according to the corresponding prediction results. The n training set groups are the pre-groupings of the training set. Furthermore, the grouping lower bound of each training set group can be obtained. The grouping lower bound is the minimum value of the corresponding prediction result in each group. The pre-groupings are sorted according to the size of the grouping lower bound to obtain sorted pre-groupings. The sorting rule can be from small to large, and the specific sorting rule is not limited in this application. Then, based on the sorted n pre-groups, the number of sample data in each training set pre-group is counted. Then, based on the sorted n pre-groups and the number of sample data in each training set pre-group, centralized grouping is performed to obtain the grouped training set.

[0079] The solution of the above embodiment, by sorting the prediction results and combining them with the preset inter-group distance, obtains the pre-grouping of the training set, obtains the number of sample data in each pre-group based on the pre-grouping, and obtains the grouped training set through centralized grouping based on the pre-grouping and the number of sample data in each pre-group. The grouped training set can be used to train the initial second machine learning model, providing a prerequisite for obtaining an explainable machine learning model for the business scenario.

[0080] In one embodiment, Figure 4 As shown, S460 includes:

[0081] S462: If the number of pre-grouped sample data is less than the number of adjacent pre-grouped samples in a preset ratio, merging the pre-grouped sample data with the corresponding adjacent pre-grouped sample data;

[0082] S464: If the number of pre-grouped sample data is greater than or equal to the number of adjacent pre-grouped samples in a preset ratio, a grouped training set is obtained based on the pre-grouped sample data.

[0083] In this embodiment, if the number of sample data in a pre-group is less than the number of samples in an adjacent pre-group in a preset ratio, the sample data in the pre-group is merged with the sample data in the corresponding adjacent pre-group. Specifically, the number of sample data in the pre-group is judged in turn. If the number of sample data in a certain pre-group is less than the number of samples in an adjacent pre-group in a preset ratio, the first pre-group adjacent to the pre-group is the adjacent pre-group. The size of the preset ratio is not specifically limited in this application. The sample data of the pre-group and the sample data of the adjacent pre-group are merged into one pre-group. After one or more merging operations, if the number of sample data in the pre-group is greater than or equal to the number of sample data in the adjacent pre-group in a preset ratio, the merging is stopped to obtain N merged training set pre-groups, which is the grouped training set.

[0084] The solution of the above embodiment is to judge the size of the number of pre-grouped sample data and the number of adjacent pre-grouped samples in a preset ratio. If it is less than the number of adjacent pre-grouped samples in the preset ratio, the pre-grouped and corresponding adjacent pre-grouped sample data are merged. If it is greater than or equal to the number of adjacent pre-grouped samples in the preset ratio, a grouped training set is obtained based on the pre-grouped sample data. The grouped training set can be used to train the initial second machine learning model, which provides a prerequisite for obtaining an interpretable machine learning model for the business scenario.

[0085] In one embodiment, multiple second machine learning models of business scenarios are aggregated to obtain an interpretable machine learning model group of the business scenarios, and then the method also includes: interpreting the interpretable machine learning model group of the business scenarios to obtain the interpretation results of the interpretable machine learning model group of the business scenarios.

[0086] In this embodiment, by interpreting multiple machine learning models in the interpretable machine learning model group of the business scenario separately, the interpretation results of multiple machine learning models are obtained, that is, the interpretation results of the interpretable machine learning model group of the business scenario.

[0087] The solution of the above embodiment, by interpreting the interpretable machine learning model group of the business scenario, obtains the interpretation results of the interpretable machine learning model group of the business scenario, which can support good interpretability results.

[0088] In one embodiment, Figure 5 As shown, the above target object risk level prediction method also includes:

[0089] S720: Obtain sample data of the business scenario to be tested;

[0090] S740: Predicting the sample data of the business scenario to be tested using a first machine learning model of the business scenario to obtain a first prediction result of the business scenario to be tested;

[0091] S760: Obtain a second prediction result of the business scenario to be tested based on the first prediction result of the business scenario to be tested and the explainable machine learning model group.

[0092] In this embodiment, sample data of the business scenario to be tested is obtained, and the sample data of the business scenario to be tested is predicted by the first machine learning model of the business scenario to obtain a first prediction result of the business scenario to be tested. Specifically, in the actual business scenario, sample data of the actual business scenario is obtained, that is, sample data of the business scenario to be tested, and the first machine learning model of the business scenario is used to predict the sample data of the business scenario to be tested to obtain a prediction value, which is the first prediction result of the business scenario to be tested. Then, based on the first prediction result of the business scenario to be tested and the explainable machine learning model group, the second prediction result of the business scenario to be tested is obtained.

[0093] The solution of the above embodiment obtains sample data of the business scenario to be tested, predicts the sample data of the business scenario to be tested through the first machine learning model of the business scenario, obtains a first prediction result of the business scenario to be tested, and obtains a second prediction result of the business scenario to be tested based on the first prediction result of the business scenario to be tested and the interpretable machine learning model group. This method of indirectly predicting part of the prediction results of the first machine learning model of the business scenario through the interpretable machine learning model group reduces the learning difficulty of the interpretable machine learning model.

[0094] In one embodiment, Figure 6 As shown, S760 includes:

[0095] S762: Obtaining a grouping space based on the grouped training set;

[0096] S764: Determine the grouping space corresponding to the first prediction result to obtain a target grouping space;

[0097] S766: Selecting a second machine learning model corresponding to the target grouping space through the interpretable machine learning model group;

[0098] S768: Predict the sample data of the business scenario to be tested through the selected second machine learning model to obtain a second prediction result of the business scenario to be tested.

[0099] In this embodiment, a grouping space is obtained based on the grouped training set, and each grouping space includes a grouped training set and the corresponding prediction result of the training set. According to the prediction result range in the grouping space, it is determined that the size of the first prediction result belongs to the prediction result range of which grouping space, and the molecular space corresponding to the first prediction result is obtained, that is, the target grouping space. According to the grouped training set corresponding to the target grouping space, the second machine learning model of the business scenario corresponding to the grouped training set is obtained from the interpretable machine learning model group. The sample data of the business scenario to be tested is predicted by the selected second machine learning model of the business scenario to obtain the second prediction result of the business scenario to be tested.

[0100] The solution of the above embodiment obtains a grouping space based on the grouped training set, determines the grouping space corresponding to the first prediction result, selects a second machine learning model corresponding to the target grouping space, and predicts the sample data of the business scenario to be tested to obtain the second prediction result of the business scenario to be tested. This method of indirectly predicting part of the prediction results of the first machine learning model of the business scenario through an interpretable machine learning model group reduces the learning difficulty of the interpretable machine learning model.

[0101] To illustrate the target object risk level prediction method and effect in this solution in detail, the following is a most detailed example:

[0102] For the common risk prevention and control business scenarios in the financial field, the financial status and credit status data of the target objects in the risk prevention and control business scenarios are obtained, 80% of the sample data are selected as the training set, and the remaining sample data are used as the test set. The initial first machine learning model is trained according to the training set to obtain the first machine learning model of the risk prevention and control business scenario. The initial first machine learning model is a high-precision machine learning model lightGBM. The first machine learning model of the risk prevention and control business scenario is tested according to the test set to obtain the prediction results. The method of grouping according to the prediction results is adopted. Specifically, the prediction results are sorted, and the pre-grouping of the training set is obtained based on the preset inter-group distance. According to the pre-grouping, the number of sample data of each pre-group is obtained. If the number of sample data of the pre-group is less than the number of samples of the adjacent pre-group of the preset ratio, the pre-grouped sample data is merged with the corresponding adjacent pre-grouped sample data. If the number of sample data of the pre-group is greater than If the number of adjacent pre-grouped samples is equal to the preset ratio, then a grouped training set is obtained according to the pre-grouped sample data, and multiple initial second machine learning models are trained according to the grouped training set to obtain multiple second machine learning models for the risk prevention and control business scenario. The initial second machine learning model is a highly interpretable machine learning model decision tree, and multiple second machine learning models for the risk prevention and control business scenario are aggregated to obtain an interpretable machine learning model group for the risk prevention and control business scenario. The interpretable machine learning model group for the risk prevention and control business scenario is interpreted to obtain the interpretation results of the interpretable machine learning model group for the risk prevention and control business scenario. According to the interpretation results of the interpretable machine learning model group for the risk prevention and control business scenario, the specific reasons affecting the risk level of the target object can be obtained. If the risk level of the target object is high risk, a specific explanation can be given for the high risk level of the target object based on whether the financial status and credit status data of the target object are good.

[0103] Obtain sample data of the risk prevention and control business scenario to be tested; predict the sample data of the risk prevention and control business scenario to be tested through the first machine learning model of the risk prevention and control business scenario to obtain a first prediction result of the risk prevention and control business scenario to be tested; obtain the grouping space according to the grouped training set, determine the grouping space corresponding to the first prediction result, and obtain the target grouping space; select the second machine learning model corresponding to the target grouping space according to the interpretable machine learning model group; predict the sample data of the risk prevention and control business scenario to be tested through the selected second machine learning model to obtain a second prediction result of the risk prevention and control business scenario to be tested; if the second prediction result shows that the risk level of the target object is low risk, the corresponding explanation result is that the credit status of the target object is good and the debt is high, and the bank account manager can recommend suitable wealth management products to the target object based on the risk status prediction result of the target object.

[0104] The method of the present application can also be used for smart marketing business scenarios in the financial field. Based on the investment preferences and financial status data of the target object, an interpretable machine learning model for the smart marketing business scenario is constructed. The bank account manager obtains the prediction results of the interpretable machine learning model for the smart marketing business scenario for the target object and obtains a specific explanation of the prediction results. The account manager recommends suitable financial products to the target object based on the prediction results and the explanation results. It should be noted that the target object risk level prediction method and device disclosed in this application can also be used for target object risk level prediction in other business scenarios, and the business scenarios to which the target object risk level prediction method and device disclosed in this application are applicable are not limited.

[0105] The solution of the above embodiment comprehensively uses an initial first machine learning model that is precision-oriented and an initial second machine learning model that is interpretability-oriented. Specifically, the initial first machine learning model is trained with a training set, and based on the prediction results of the training set, a grouped training set is obtained by centralized grouping; then, multiple initial second machine learning models are trained with the grouped training set, thereby obtaining multiple second machine learning models with excellent interpretability, and finally constructing an interpretable machine learning model group for business scenarios, which can support obtaining good interpretability results.

[0106] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed 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 portion of steps or stages in other steps.

[0107] Based on the same inventive concept, embodiments of the present application also provide a target object risk level prediction device for implementing the target object risk level prediction method described above. The solution to the problem provided by this device is similar to the solution described in the method described above. Therefore, the specific limitations of one or more target object risk level prediction device embodiments provided below can be found in the above-mentioned limitations of the target object risk level prediction method and will not be repeated here.

[0108] In one embodiment, Figure 7As shown, a target object risk level prediction device 800 is provided, comprising: a data acquisition module 810, a first machine learning model acquisition module 820, a prediction result acquisition module 830, a training set grouping module 840, a second machine learning model acquisition module 850 and a second machine learning model aggregation module 860, wherein:

[0109] The data acquisition module 810 is used to acquire sample data of a business scenario, where the sample data includes a training set and a test set.

[0110] The first machine learning model acquisition module 820 is used to train the initial first machine learning model according to the training set to obtain the first machine learning model of the business scenario. The initial first machine learning model is a precision-oriented machine learning model.

[0111] The prediction result acquisition module 830 is used to test the first machine learning model of the business scenario according to the test set to obtain the prediction result.

[0112] The training set grouping module 840 is used to group the training set according to the prediction results to obtain the grouped training set.

[0113] The second machine learning model acquisition module 850 is used to train multiple initial second machine learning models according to the grouped training sets to obtain multiple second machine learning models for business scenarios. The initial second machine learning model is an explainability-oriented machine learning model.

[0114] The second machine learning model collection module 860 is used to collect multiple second machine learning models of business scenarios to obtain an explainable machine learning model group of the business scenarios.

[0115] The above-mentioned target object risk level prediction device comprehensively uses an initial first machine learning model oriented towards accuracy and an initial second machine learning model oriented towards interpretability. Specifically, the initial first machine learning model is trained with a training set, and according to the prediction results of the training set, the grouped training set is obtained through centralized grouping; the grouped training set is then used to train multiple initial second machine learning models, thereby obtaining multiple second machine learning models with excellent interpretability, and finally constructing an interpretable machine learning model group for business scenarios, which can support obtaining good interpretability results.

[0116] In one embodiment, the training set grouping module 840 is also used to sort the prediction results, and obtain the pre-grouping of the training set in combination with the preset inter-group distance; based on the pre-grouping, obtain the number of sample data in each pre-grouping; and based on the pre-grouping and the number of sample data in each pre-grouping, obtain the grouped training set through centralized grouping.

[0117] In one embodiment, the training set grouping module 840 is also used to merge the pre-grouped sample data with the corresponding adjacent pre-grouped sample data if the number of pre-grouped sample data is less than the preset ratio of the number of adjacent pre-grouped samples; if the number of pre-grouped sample data is greater than or equal to the preset ratio of the number of adjacent pre-grouped samples, obtain the grouped training set based on the pre-grouped sample data.

[0118] In one embodiment, the second machine learning model collection module 860 is also used to interpret the interpretable machine learning model group of the business scenario to obtain the interpretation results of the interpretable machine learning model group of the business scenario.

[0119] In one embodiment, the target object risk level prediction device 800 is also used to obtain sample data of the business scenario to be tested; predict the sample data of the business scenario to be tested through the first machine learning model of the business scenario to obtain a first prediction result of the business scenario to be tested; and obtain a second prediction result of the business scenario to be tested based on the first prediction result of the business scenario to be tested and the explainable machine learning model group.

[0120] In one embodiment, the target object risk level prediction device 800 is also used to obtain a grouping space based on the grouped training set; determine the grouping space corresponding to the first prediction result to obtain a target grouping space; select a second machine learning model corresponding to the target grouping space through an interpretable machine learning model group; and predict the sample data of the business scenario to be tested through the selected second machine learning model to obtain a second prediction result of the business scenario to be tested.

[0121] Each module in the target object risk level prediction device described above may be implemented in whole or in part through software, hardware, or a combination thereof. Each module may be embedded in or independent of a processor in a computer device in hardware form, or may be stored in a computer device memory in software form, so that the processor can call and execute the corresponding operations of each module.

[0122] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 8As shown. The computer device includes a processor, a memory and a network interface connected via a system bus. 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 the interpretation results of the interpretable machine learning model group of the business scenario and the second prediction result of the business scenario to be tested. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for predicting the risk level of a target object is implemented.

[0123] Those skilled in the art will understand that Figure 8 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 shown in the figure, or combine certain components, or have a different component arrangement.

[0124] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0125] Obtain sample data for business scenarios, including training sets and test sets;

[0126] Training the initial first machine learning model according to the training set to obtain a first machine learning model for the business scenario, where the initial first machine learning model is a precision-oriented machine learning model;

[0127] Test the first machine learning model of the business scenario based on the test set to obtain prediction results;

[0128] According to the prediction results, the training set is grouped to obtain the grouped training set;

[0129] Training multiple initial second machine learning models according to the grouped training sets to obtain multiple second machine learning models for business scenarios, where the initial second machine learning models are interpretability-oriented machine learning models;

[0130] Aggregate multiple second machine learning models of business scenarios to obtain an explainable machine learning model group of the business scenarios.

[0131] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0132] The prediction results are sorted and combined with the preset inter-group distance to obtain the pre-grouping of the training set; based on the pre-grouping, the number of sample data in each pre-grouping is obtained; based on the pre-grouping and the number of sample data in each pre-grouping, the grouped training set is obtained through centralized grouping.

[0133] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0134] If the number of pre-grouped sample data is less than the preset ratio of the number of adjacent pre-grouped samples, the pre-grouped sample data is merged with the corresponding adjacent pre-grouped sample data; if the number of pre-grouped sample data is greater than or equal to the preset ratio of the number of adjacent pre-grouped samples, the grouped training set is obtained based on the pre-grouped sample data.

[0135] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0136] Explain the interpretable machine learning model group of the business scenario and obtain the interpretation results of the interpretable machine learning model group of the business scenario.

[0137] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0138] Obtain sample data of the business scenario to be tested; predict the sample data of the business scenario to be tested through the first machine learning model of the business scenario to obtain a first prediction result of the business scenario to be tested; obtain a second prediction result of the business scenario to be tested based on the first prediction result of the business scenario to be tested and the explainable machine learning model group.

[0139] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0140] According to the grouped training set, a grouping space is obtained; the grouping space corresponding to the first prediction result is determined to obtain the target grouping space; through the interpretable machine learning model group, a second machine learning model corresponding to the target grouping space is selected; through the selected second machine learning model, the sample data of the business scenario to be tested is predicted to obtain a second prediction result of the business scenario to be tested.

[0141] 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 following steps are implemented:

[0142] Obtain sample data for business scenarios, including training sets and test sets;

[0143] Training the initial first machine learning model according to the training set to obtain a first machine learning model for the business scenario, where the initial first machine learning model is a precision-oriented machine learning model;

[0144] Test the first machine learning model of the business scenario based on the test set to obtain prediction results;

[0145] According to the prediction results, the training set is grouped to obtain the grouped training set;

[0146] Training multiple initial second machine learning models according to the grouped training sets to obtain multiple second machine learning models for business scenarios, where the initial second machine learning models are interpretability-oriented machine learning models;

[0147] Aggregate multiple second machine learning models of business scenarios to obtain an explainable machine learning model group of the business scenarios.

[0148] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0149] The prediction results are sorted and combined with the preset inter-group distance to obtain the pre-grouping of the training set; based on the pre-grouping, the number of sample data in each pre-grouping is obtained; based on the pre-grouping and the number of sample data in each pre-grouping, the grouped training set is obtained through centralized grouping.

[0150] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0151] If the number of pre-grouped sample data is less than the preset ratio of the number of adjacent pre-grouped samples, the pre-grouped sample data is merged with the corresponding adjacent pre-grouped sample data; if the number of pre-grouped sample data is greater than or equal to the preset ratio of the number of adjacent pre-grouped samples, the grouped training set is obtained based on the pre-grouped sample data.

[0152] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0153] Explain the interpretable machine learning model group of the business scenario and obtain the interpretation results of the interpretable machine learning model group of the business scenario.

[0154] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0155] Obtain sample data of the business scenario to be tested; predict the sample data of the business scenario to be tested through the first machine learning model of the business scenario to obtain a first prediction result of the business scenario to be tested; obtain a second prediction result of the business scenario to be tested based on the first prediction result of the business scenario to be tested and the explainable machine learning model group.

[0156] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0157] According to the grouped training set, a grouping space is obtained; the grouping space corresponding to the first prediction result is determined to obtain the target grouping space; through the interpretable machine learning model group, a second machine learning model corresponding to the target grouping space is selected; through the selected second machine learning model, the sample data of the business scenario to be tested is predicted to obtain a second prediction result of the business scenario to be tested.

[0158] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0159] Obtain sample data for business scenarios, including training sets and test sets;

[0160] Training the initial first machine learning model according to the training set to obtain a first machine learning model for the business scenario, where the initial first machine learning model is a precision-oriented machine learning model;

[0161] Test the first machine learning model of the business scenario based on the test set to obtain prediction results;

[0162] According to the prediction results, the training set is grouped to obtain the grouped training set;

[0163] Training multiple initial second machine learning models according to the grouped training sets to obtain multiple second machine learning models for business scenarios, where the initial second machine learning models are interpretability-oriented machine learning models;

[0164] Aggregate multiple second machine learning models of business scenarios to obtain an explainable machine learning model group of the business scenarios.

[0165] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0166] The prediction results are sorted and combined with the preset inter-group distance to obtain the pre-grouping of the training set; based on the pre-grouping, the number of sample data in each pre-grouping is obtained; based on the pre-grouping and the number of sample data in each pre-grouping, the grouped training set is obtained through centralized grouping.

[0167] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0168] If the number of pre-grouped sample data is less than the preset ratio of the number of adjacent pre-grouped samples, the pre-grouped sample data is merged with the corresponding adjacent pre-grouped sample data; if the number of pre-grouped sample data is greater than or equal to the preset ratio of the number of adjacent pre-grouped samples, the grouped training set is obtained based on the pre-grouped sample data.

[0169] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0170] Explain the interpretable machine learning model group of the business scenario and obtain the interpretation results of the interpretable machine learning model group of the business scenario.

[0171] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0172] Obtain sample data of the business scenario to be tested; predict the sample data of the business scenario to be tested through the first machine learning model of the business scenario to obtain a first prediction result of the business scenario to be tested; obtain a second prediction result of the business scenario to be tested based on the first prediction result of the business scenario to be tested and the explainable machine learning model group.

[0173] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0174] According to the grouped training set, a grouping space is obtained; the grouping space corresponding to the first prediction result is determined to obtain the target grouping space; through the interpretable machine learning model group, a second machine learning model corresponding to the target grouping space is selected; through the selected second machine learning model, the sample data of the business scenario to be tested is predicted to obtain a second prediction result of the business scenario to be tested.

[0175] It should be noted that the target object risk level prediction method and device disclosed in this application can be used for target object risk level prediction in the financial field, and can also be used for target object risk level prediction in any field other than the financial field. The application field of the target object risk level prediction method and device disclosed in this application is not limited. 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.

[0176] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented 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 memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may 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 may include random access memory (RAM) or external cache memory, etc. By way of 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 the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in the various embodiments provided herein may be, but are not limited to, a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, and the like.

[0177] The technical features of the above embodiments can 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 specification.

[0178] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for predicting the risk level of a target object, characterized in that: The method comprises: Follow these steps on your computer: Obtaining sample data of a business scenario, the sample data including a training set and a test set, wherein the business scenario is a risk prevention and control business scenario, obtaining financial status and credit status data of a target object in the risk prevention and control business scenario, and obtaining the training set and the test set; Training an initial first machine learning model according to the training set to obtain a first machine learning model for the business scenario, wherein the initial first machine learning model is a precision-oriented machine learning model; Testing the first machine learning model of the business scenario according to the test set to obtain a prediction result; Grouping the training set according to the prediction result to obtain a grouped training set; Training multiple initial second machine learning models according to the grouped training sets to obtain multiple second machine learning models for the business scenario, where the initial second machine learning models are interpretability-oriented machine learning models; Aggregating multiple second machine learning models of the business scenario to obtain an explainable machine learning model group of the business scenario; Interpreting the interpretable machine learning model group for the risk prevention and control business scenario to obtain an interpretation result of the interpretable machine learning model group for the risk prevention and control business scenario, and deriving specific reasons affecting the risk level of the target object based on the interpretation result; The method further comprises: Obtain sample data for the business scenario to be tested; Predicting the sample data of the business scenario to be tested by using the first machine learning model of the business scenario to obtain a first prediction result of the business scenario to be tested; Obtaining a grouping space according to the grouped training set; Determine a grouping space corresponding to the first prediction result to obtain a target grouping space; Selecting the second machine learning model corresponding to the target grouping space according to the interpretable machine learning model group; Predicting the sample data of the business scenario to be tested using the selected second machine learning model to obtain a second prediction result of the business scenario to be tested; Each grouping space includes a grouped training set and the prediction result of the corresponding training set.

2. The target object risk level prediction method according to claim 1, characterized in that: The training set is grouped according to the prediction result, and the obtained grouped training set includes: Sorting the prediction results and combining them with the preset inter-group distance to obtain the pre-grouping of the training set; According to the pre-grouping, obtaining the number of sample data in each pre-group; According to the pre-grouping and the number of sample data in each pre-grouping, a grouped training set is obtained by centralized grouping.

3. The target object risk level prediction method according to claim 2, characterized in that: According to the pre-grouping and the number of sample data in each pre-grouping, the grouped training set obtained by centralized grouping includes: If the number of the pre-grouped sample data is less than the number of adjacent pre-grouped samples in a preset ratio, merging the pre-grouped sample data with the corresponding adjacent pre-grouped sample data; If the number of the pre-grouped sample data is greater than or equal to the number of adjacent pre-grouped samples in a preset ratio, a grouped training set is obtained based on the pre-grouped sample data.

4. A target object risk level prediction device, characterized in that: Applied to computer equipment, the device comprises: a data acquisition module, configured to acquire sample data for a business scenario, the sample data including a training set and a test set, wherein the business scenario is a risk prevention and control business scenario, and to acquire financial status and credit status data of a target subject in the risk prevention and control business scenario to obtain the training set and the test set; A first machine learning model acquisition module, configured to train an initial first machine learning model according to the training set to obtain a first machine learning model for the business scenario, wherein the initial first machine learning model is a precision-oriented machine learning model; A prediction result acquisition module, configured to test the first machine learning model of the business scenario according to the test set to obtain a prediction result; A training set grouping module, configured to group the training set according to the prediction results to obtain grouped training sets; a second machine learning model acquisition module, configured to train a plurality of initial second machine learning models according to the grouped training sets to obtain a plurality of second machine learning models for the business scenario, wherein the initial second machine learning models are interpretability-oriented machine learning models; a second machine learning model aggregation module, configured to aggregate multiple second machine learning models of the business scenario to obtain an interpretable machine learning model group of the business scenario, interpret the interpretable machine learning model group of the risk prevention and control business scenario, obtain an interpretation result of the interpretable machine learning model group of the risk prevention and control business scenario, and derive specific reasons affecting the risk level of the target object based on the interpretation result; The device is further configured to obtain sample data of a business scenario to be tested; predict the sample data of the business scenario to be tested using a first machine learning model of the business scenario to obtain a first prediction result of the business scenario to be tested; obtain a grouping space based on the grouped training set; determine the grouping space corresponding to the first prediction result to obtain a target grouping space; select a second machine learning model corresponding to the target grouping space based on an interpretable machine learning model group; predict the sample data of the business scenario to be tested using the selected second machine learning model to obtain a second prediction result of the business scenario to be tested; Each grouping space includes a grouped training set and the prediction result of the corresponding training set.

5. 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 3 are implemented.

6. 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 3 are implemented.

7. 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 3 are implemented.

Images