Model generation and use method of unbalanced sample in risk control scene and related equipment
By employing the k-fold cross-validation method in the field of financial risk control to construct a risk assessment model, the problem of sample imbalance was solved, the model's ability to identify high-risk users was improved, and the level of risk control was enhanced.
Patent Information
- Application Number
- CN202411040390.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-31
- Publication Date
- 2024-11-12
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing technologies in the field of financial risk control face the problem of imbalanced samples, which causes the model performance to decay over time, seriously weakening the level of risk control.
The k-fold cross-validation method is adopted to solve the problem of sample imbalance by constructing a risk assessment model during the training process of the basic classifier and the preset basic model, and by using the combination of multiple basic classifiers and the preset basic model.
This improved the final performance of the model, enhanced risk control, effectively identified high-risk users, and met the risk management needs of financial businesses.
Smart Images

Figure CN118941393A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of financial risk control, and in particular to a method and related equipment for generating and using a model of unbalanced samples in a risk control scenario. Background Art
[0002] From the perspective of the model, sample imbalance or imbalance means that in the model classification task, the training sample sizes of different categories (divided into positive and negative classes) vary greatly. Generally speaking, if the sample category ratio (majority class: minority class) is significantly greater than 1:1 (such as 100:1), it can be considered that there is a sample imbalance problem. In particular, in the field of financial risk control, the samples we use to train models and assess risks generally have sample imbalance problems, that is, the proportion of negative samples is particularly small. For example, the proportion of customers with a customer label of MOB6,30+ (month on book, MOB) is 6 months overdue for more than 30 days) may be less than 1%. In other words, only 100 overdue customer samples out of 10,000 customers who are granted credit can be used for model training. If you want to train a good model, such a sample size is obviously not enough.
[0003] In the prior art, the sample imbalance problem is generally handled through simple sampling strategies, such as undersampling or oversampling. This method of artificially changing the ratio of positive and negative samples will lead to a large amount of positive sample information missing in the modeling process after sampling, and ultimately cause the model effect to decay over time, seriously weakening the subsequent risk control level. Therefore, there is an urgent need for a method that can maintain the model effect and overcome the sample imbalance problem that exists during model training. Summary of the invention
[0004] In view of this, the purpose of the present invention is to provide a method and related equipment for generating and using a model of unbalanced samples in a risk control scenario, so as to overcome the problem of sample imbalance during current model training.
[0005] To achieve the above objectives, the present invention adopts the following technical solutions:
[0006] In a first aspect, an embodiment of the present application provides a method for generating and using a model of unbalanced samples in a risk control scenario, including:
[0007] Collect user data to generate original sample sets;
[0008] Constructing a risk assessment model based on the original sample set and test data generated during the training of the basic classifier and the preset basic model through k-fold cross validation;
[0009] Predicting a risk score for a user based on the risk assessment model;
[0010] The risky users are screened out according to the risk scores.
[0011] Further, in some embodiments of the present application, the risk assessment model is constructed based on the original sample set and the test data generated in the training process of the basic classifier and the preset basic model through k-fold cross validation, including:
[0012] Determine a plurality of basic classifiers, and set parameters of each of the basic classifiers;
[0013] Dividing the original sample set into an original training set and an original test set, and generating k first sub-sample sets based on the original training set by k-fold cross validation;
[0014] For each of the basic classifiers, the k first sub-sample sets are divided into k combinations, wherein each combination includes a first sub-training set and a first sub-test set;
[0015] For each of the basic classifiers, training is performed using each first sub-training set to obtain a trained basic classifier, and a first sub-test set in the same group as the first sub-training set is input into the trained basic classifier to obtain first prediction data;
[0016] The original test set is input into the trained basic classifier to obtain second prediction data.
[0017] Further, in some embodiments of the present application, for each of the basic classifiers, dividing the k first sub-sample sets into k combinations includes:
[0018] One of the k first sub-sample sets is sequentially used as a first sub-test set, and the other first sub-sample sets except the first sub-sample set are used as first sub-training sets in the same group as the first sub-test set.
[0019] Further, in some embodiments of the present application, the risk assessment model is constructed based on the original sample set and the test data generated in the training process of the basic classifier and the preset basic model through k-fold cross validation, and further includes:
[0020] Using the first prediction data as a loop training set and using the second prediction data as a loop test set;
[0021] Generate k second sub-sample sets based on the cyclic training set by k-fold cross validation, wherein each second sub-sample set includes a second sub-training set and a second sub-test set;
[0022] For each of the preset basic models, training is performed through each second sub-training set to obtain a trained preset basic model, and a second sub-test set corresponding to the second sub-training set is input into the trained preset basic model, and the output result is used to update the cyclic training data;
[0023] Input the loop test set into the trained preset basic model, and update the loop training data with the output result;
[0024] Retrain and update the cyclic training set and the cyclic test set based on the new cyclic training set and the cyclic test set until the basic model indicators meet the standards.
[0025] Furthermore, in some embodiments of the present application, the indicators include a model AUC value and a model KS value.
[0026] Furthermore, in some embodiments of the present application, the basic classifier includes at least one of XGBoost, Adaboost, GBDT, RF, SVM, and LR.
[0027] Furthermore, in some embodiments of the present application, the basic model includes an RF classifier.
[0028] Furthermore, in some embodiments of the present application, the user data includes basic customer information, customer loan history behavior, customer Internet behavior and customer online payment behavior.
[0029] In a second aspect, the embodiment of the present application further provides a model generation and use device for unbalanced samples in a risk control scenario, including:
[0030] A collection module, used to collect user data to generate an original sample set;
[0031] A construction module, for constructing a risk assessment model based on the original sample set and test data generated in the training process of the basic classifier and the preset basic model through k-fold cross validation;
[0032] A prediction model, used to predict a user's risk score based on the risk assessment model;
[0033] The screening module is used to screen out risky users according to the risk scores.
[0034] In a third aspect, an embodiment of the present application further provides a model generation and use device for unbalanced samples in a risk control scenario, characterized in that it includes a processor and a memory, wherein the processor is connected to the memory:
[0035] Wherein, the processor is used to call and execute the program stored in the memory;
[0036] The memory is used to store the program, and the program is at least used to execute the model generation and use method of unbalanced samples in the risk control scenario described in any one of claims 1-8.
[0037] The present invention relates to the technical field of financial risk control, and specifically to a method and related equipment for generating and using a model of unbalanced samples in a risk control scenario, the method comprising first collecting user data to generate an original sample set; then building a risk assessment model based on the original sample set and test data generated in the training process of a basic classifier and a preset basic model through k-fold cross validation; then predicting the risk score of the user according to the risk assessment model; and finally screening out risky users according to the risk score. In this way, by using the test data generated in the training process of a basic classifier and a preset basic model through k-fold cross validation for model training and testing, the problem of sample imbalance existing in the establishment of a model in the technical field of financial risk control can be solved, and the final effect of the model can be greatly improved, thereby improving the level of risk control. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0039] Figure 1 It is a flowchart of a method for generating and using a model of unbalanced samples in a risk control scenario provided by an embodiment of the present invention;
[0040] Figure 2 It is a flowchart of building a model in the method for generating and using a model of unbalanced samples in a risk control scenario provided by an embodiment of the present invention;
[0041] Figure 3 It is a structural schematic diagram of a device for generating and using a model of unbalanced samples in a risk control scenario provided by an embodiment of the present invention;
[0042] Figure 4 It is a structural schematic diagram of a device for generating and using a model of unbalanced samples in a risk control scenario of the present invention. DETAILED DESCRIPTION
[0043] To make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described in detail below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.
[0044] Figure 1 This is a flow chart of the method for generating and using a model of unbalanced samples in a risk control scenario provided by an embodiment of the present invention. Figure 1 , this embodiment may include the following steps:
[0045] S101: Collect user data to generate an original sample set.
[0046] Specifically, user data may include basic customer information, customer loan history, customer Internet behavior, and customer online payment behavior.
[0047] In practical applications, the user data may specifically be the data shown in Table 1 below:
[0048] Table 1 User data
[0049]
[0050]
[0051] S102: construct a risk assessment model based on the original sample set and the test data generated during the training of the basic classifier and the preset basic model through k-fold cross validation.
[0052] Specifically, in the present application, firstly, the basic classifier is trained by k-fold cross validation based on the original sample set (the process includes dividing the original sample set into the original training set and the original test set for training and testing), and the test data is obtained. Then, based on the obtained test data, the preset basic model is cyclically trained and tested (the above test data is updated based on the effect of the trained model to continue training and testing), and the final model, i.e., the risk assessment model, is obtained. It should be noted that the above original sample set is an unbalanced sample set of positive and negative samples.
[0053] S103. Predict the user's risk score based on the risk assessment model.
[0054] S104. Screen out risky users based on risk scores.
[0055] Specifically, after obtaining the risk assessment model, the data of the users who need to be scored are input into the value model to obtain the risk scores of all users who need risk scores, and then the risky users or high-risk users are determined according to the risk scores and the preset thresholds.
[0056] The model generation and use method of unbalanced samples in the risk control scenario provided by the present invention first collects user data to generate an original sample set; then, based on the original sample set and the test data generated by k-fold cross-validation in the training process of the basic classifier and the preset basic model, a risk assessment model is constructed; then, the risk score of the user is predicted according to the risk assessment model; finally, risky users are screened out according to the risk score. In this way, by using the test data generated by k-fold cross-validation in the training process of the basic classifier and the preset basic model for model training and testing, the problem of sample imbalance existing in the establishment of models in the field of financial risk control technology can be solved, and the final effect of the model can be greatly improved, thereby improving the level of risk control.
[0057] Furthermore, in some embodiments of the present application, a risk assessment model is constructed based on the original sample set and the test data generated in the training process of the basic classifier and the preset basic model through k-fold cross validation, which may specifically include the following steps:
[0058] S201: Determine multiple basic classifiers and set parameters of each basic classifier.
[0059] Specifically, multiple basic classifiers are first selected, such as at least one of XGBoost, Adaboost, GBDT, RF, SVM, and LR, and then the model parameters corresponding to each basic classifier are set.
[0060] S202 , dividing the original sample set into an original training set and an original test set, and generating k first sub-sample sets based on the original training set through k-fold cross validation.
[0061] S203 . For each basic classifier, generate k combinations based on the k first sub-sample sets, wherein each combination includes a first sub-training set and a first sub-test set.
[0062] Specifically, the original data set is first divided into an original training set and an original test set, and then k-fold cross validation is performed on the original training set to obtain k sub-sample sets, namely k first sub-sample sets. Then, for each i-th base classifier, the k sub-sample sets are divided into k pairs of combinations of the first sub-test set and the first sub-training set, where each of the k combinations includes a first sub-test set and a first sub-training set.
[0063] Specifically, one first sub-sample set is selected in turn as the first sub-test set, and the remaining k-1 first sub-sample sets are selected as sub-training sets to form k pairs of combinations of the first sub-test set and the first sub-training set.
[0064] S204, for each basic classifier, training is performed using each first sub-training set to obtain a trained basic classifier, and a first sub-test set in the same group as the first sub-training set is input into the trained basic classifier to obtain first prediction data.
[0065] S205 , input the original test set into the trained basic classifier to obtain second prediction data.
[0066] Specifically, based on each first sub-training set, each i-th basic classifier is trained, and each trained basic classifier predicts the first sub-test set corresponding to the first sub-training set trained by it, and obtains a sub-prediction data a and b respectively. i1 ,a i2 …a ik , the original test set is predicted by each trained basic classifier to obtain b sub-prediction data b i2 ,b i2 …b ik Then the above a prediction data a i1 ,a i2 …a ik Combine to generate A prediction data A i (i.e. the first prediction data), and the b sub-prediction data b i1 ,b i2 …b ik Combine and generate prediction data B i (ie, the second prediction data); wherein, i=1,2…n, and n is the number of basic classifiers.
[0067] On this basis, the process of training the preset basic model based on the first prediction data and the second prediction data to obtain the final model may include the following steps:
[0068] S301: Use the first prediction data as a cyclic training set, and use the second prediction data as a cyclic testing set.
[0069] S302 : Generate k second sub-sample sets based on the cyclic training set through k-fold cross validation, wherein each second sub-sample set includes a second sub-training set and a second sub-test set.
[0070] S303: For each preset basic model, training is performed through each second sub-training set to obtain a trained preset basic model, and a second sub-test set corresponding to the second sub-training set is input into the trained preset basic model, and the output result is used to update the cyclic training data.
[0071] S304: Input the loop test set into the trained preset basic model, and use the output result to update the loop training data.
[0072] S305: Retrain and update the cyclic training set and cyclic test set based on the new cyclic training set and cyclic test set until the basic model indicators meet the standards.
[0073] Specifically, the A prediction data, i.e., A1, A2, ... A in the first prediction data, are converted into n Combined as a training set as a cyclic training set, the B prediction data, that is, the second prediction data B1, B2...B n The combined test set is the cyclic test set. The above cyclic training set is subjected to k-fold cross validation to obtain k second sub-sample sets, and the k second sub-sample sets are trained (same as the k-fold cross validation principle in the above process, also divided into k combinations, each combination includes a second sub-training set and a second sub-test set) with each preset basic model (in practice, in order to ensure the effect of the final model, multiple preset basic models are used for training and testing during training and testing, and the best effect is finally obtained or multiple models are integrated to obtain the final risk assessment model), and the trained preset basic model is used to predict the second sub-test set corresponding to the second sub-training set trained with it, and obtain A prediction data A1, A2…A m , to update the above cycle training set, and use the trained preset basic model to predict the cycle test set to obtain B prediction data B1, B2...B m , to update the cycle test set, where m is the preset number of models.
[0074] On this basis, the above step S305 is looped until the AUC index of the model reaches the AUC preset value and the KS index reaches the KS preset value, thereby obtaining the final risk assessment model.
[0075] In some embodiments of the present application, the above-mentioned basic classifier may include at least one of XGBoost, Adaboost, GBDT, RF, SVM, LR, and of course other classifiers, which can be changed according to actual needs; the basic model includes the RF classifier, and similarly it can be other classifiers, which can be changed according to actual needs.
[0076] The model generation and use method of unbalanced samples in risk control scenarios provided by the present invention can perform model training on unbalanced samples in the model generation stage, and can make the AUC>0.6 and KS>0.2 of the risk control model by setting preset values of evaluation indicators, so as to develop a risk control model that meets business needs for certain specific risk control business scenarios (such as pre-loan anti-fraud, loan risk warning, etc.), and can also use the risk control model to predict customer risk scores, ultimately effectively supporting the development of financial business and risk management, and improving the level of risk control.
[0077] Figure 2 This is a flow chart of building a model in the method for generating and using a model of unbalanced samples in a risk control scenario provided by an embodiment of the present invention. Figure 2 , for an overall introduction to the construction process of the risk assessment model in this application, please refer to Figure 2 , this embodiment may include the following steps:
[0078] First, a basic classifier is selected and parameters are set; then, the original sample set generated by the initially collected user data is trained and tested using k-fold cross validation to train each basic classifier, and the first prediction data and the second prediction data are obtained; then, the first prediction data is used as a cyclic training set, and the second prediction data is used as a cyclic test set, and k-fold cross validation is used to train and test the preset basic model to obtain new test data to update the cyclic training set and the cyclic test set. At this time, it is detected whether the model index reaches the preset value. If it reaches, the final risk assessment model is obtained. If not, this step is repeated, and the updated cyclic training set and cyclic test set are used for training and testing, and then a new cyclic training set and cyclic test set are obtained, and the model effect is verified until the model index reaches the preset value.
[0079] The present invention aims at the sample imbalance that often occurs in user data in the field of credit risk control. By collecting and organizing user data as an original sample set, a risk scoring model is constructed based on the original sample set. Specifically, different types of classifiers are used in the process of building the model to perform appropriate combination or stacking on the imbalanced samples, and finally a composite hierarchical model structure is formed, and relatively limited negative samples are utilized in models at each level. Although the classifier of each layer is only a weak classifier, the integrated and stacked classifiers can effectively perform classification training on negative samples, thereby achieving the purpose of identifying bad customers, thereby improving the risk control level.
[0080] Based on the same inventive concept, the present invention also provides a model generation and use device for unbalanced samples in a risk control scenario, which is used to implement the above method embodiment. Figure 3 is a schematic diagram of the structure of a device for generating and using a model of unbalanced samples in a risk control scenario provided by an embodiment of the present invention, such as Figure 3 As shown, the device may specifically include:
[0081] A collection module 31 is used to collect user data to generate an original sample set;
[0082] A construction module 32, for constructing a risk assessment model based on the original sample set and test data generated in the training process of the basic classifier and the preset basic model through k-fold cross validation;
[0083] Prediction model 33, used to predict the risk score of the user according to the risk assessment model;
[0084] The screening module 34 is used to screen out risky users according to the risk scores.
[0085] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.
[0086] Based on the same inventive concept, the present invention also provides a model generation and use device for unbalanced samples in a risk control scenario, which is used to implement the above method embodiment. Figure 4 Schematic diagram of the structure of the device for generating and using the model of unbalanced samples in the risk control scenario of the present invention. Figure 4 As shown, the model generation and use device for unbalanced samples in the risk control scenario of this embodiment includes a processor 41 and a memory 42, and the processor 41 is connected to the memory 42. Among them, the processor 41 is used to call and execute the program stored in the memory 42; the memory 42 is used to store the program, and the program is at least used to execute the model generation and use method of unbalanced samples in the risk control scenario in the above embodiment.
[0087] The specific implementation scheme of the device for generating and using a model of unbalanced samples in a risk control scenario provided in the embodiments of the present application can refer to the implementation scheme of the method for generating and using a model of unbalanced samples in a risk control scenario in any of the above embodiments, and will not be repeated here.
[0088] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.
[0089] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" refers to at least two.
[0090] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.
[0091] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0092] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.
[0093] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.
[0094] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.
[0095] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.
[0096] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.
Claims
1. A method for generating and using a model for unbalanced samples in risk control scenarios. It is characterized in that include: Collect user data to generate original sample sets; Constructing a risk assessment model based on the original sample set and test data generated during the training of the basic classifier and the preset basic model through k-fold cross validation; Predicting a risk score for a user based on the risk assessment model; The risky users are screened out according to the risk scores.
2. The method for generating and using a model for unbalanced samples in a risk control scenario according to claim 1, It is characterized in that The risk assessment model is constructed based on the original sample set and the test data generated in the training process of the basic classifier and the preset basic model through k-fold cross validation, including: Determine a plurality of basic classifiers, and set parameters of each of the basic classifiers; Dividing the original sample set into an original training set and an original test set, and generating k first sub-sample sets based on the original training set by k-fold cross validation; For each of the basic classifiers, generating k combinations based on the k first sub-sample sets, wherein each combination includes a first sub-training set and a first sub-test set; For each of the basic classifiers, training is performed using each first sub-training set to obtain a trained basic classifier, and a first sub-test set in the same group as the first sub-training set is input into the trained basic classifier to obtain first prediction data; The original test set is input into the trained basic classifier to obtain second prediction data.
3. The method for generating and using a model for unbalanced samples in a risk control scenario according to claim 2, It is characterized in that For each of the basic classifiers, dividing the k first sub-sample sets into k combinations comprises: One of the k first sub-sample sets is sequentially used as a first sub-test set, and the other first sub-sample sets except the first sub-sample set are used as first sub-training sets in the same group as the first sub-test set.
4. The method for generating and using a model for unbalanced samples in a risk control scenario according to claim 2, It is characterized in that The risk assessment model is constructed based on the original sample set and the test data generated in the training process of the basic classifier and the preset basic model through k-fold cross validation, and further includes: Using the first prediction data as a loop training set and using the second prediction data as a loop test set; Generate k second sub-sample sets based on the cyclic training set by k-fold cross validation, wherein each second sub-sample set includes a second sub-training set and a second sub-test set; For each of the preset basic models, training is performed through each second sub-training set to obtain a trained preset basic model, and a second sub-test set corresponding to the second sub-training set is input into the trained preset basic model, and the output result is used to update the cyclic training data; Input the loop test set into the trained preset basic model, and update the loop training data with the output result; Retrain and update the cyclic training set and the cyclic test set based on the new cyclic training set and the cyclic test set until the basic model indicators meet the standards.
5. The method for generating and using a model for unbalanced samples in a risk control scenario according to claim 4, It is characterized in that The indicators include model AUC value and model KS value.
6. The method for generating and using a model for unbalanced samples in a risk control scenario according to claim 1, It is characterized in that The basic classifier includes at least one of XGBoost, Adaboost, GBDT, RF, SVM, and LR.
7. The method for generating and using a model for unbalanced samples in a risk control scenario according to claim 1, It is characterized in that The base model includes an RF classifier.
8. The method for generating and using a model for unbalanced samples in a risk control scenario according to claim 1, It is characterized in that The user data includes basic customer information, customer loan history, customer Internet behavior and customer online payment behavior.
9. A model generation and use device for unbalanced samples in risk control scenarios, It is characterized in that include: A collection module, used to collect user data to generate an original sample set; A construction module, for constructing a risk assessment model based on the original sample set and test data generated in the training process of the basic classifier and the preset basic model through k-fold cross validation; A prediction model, used to predict a user's risk score based on the risk assessment model; The screening module is used to screen out risky users according to the risk scores.
10. A device for generating and using a model for unbalanced samples in risk control scenarios. It is characterized in that The invention comprises a processor and a memory, wherein the processor is connected to the memory: Wherein, the processor is used to call and execute the program stored in the memory; The memory is used to store the program, and the program is at least used to execute the model generation and use method of unbalanced samples in the risk control scenario described in any one of claims 1-8.
Citation Information
Patent Citations
Risk control method and device based on biased sample and electronic equipment
CN111160733A
Credit risk control model construction system and method, risk control system and storage medium
CN111768285A
Internet financial client application fraud detection method based on AdaBoost
CN112581265A
Network loan risk prediction method
CN115860892A
