Claim risk model processing method and device, computer equipment and storage medium

Abstract

The embodiment of the invention belongs to the field of artificial intelligence, and relates to a claim risk model processing method, which comprises the following steps: acquiring a first training set containing original samples and risk tags of group insurance customers to train a first initial risk model to obtain a first risk model, and acquiring a risk prediction result of the first risk model for each group insurance customer; constructing a second training set according to the risk prediction result and the first training set to train a second initial risk model to obtain a second risk model; based on an integration algorithm and the second training set, performing integrated training on the first risk model and the second risk model to obtain an initial group insurance claim model; performing cross validation on a first risk model and a second risk model in the initial group insurance claim model to adjust model weights of the first risk model and the second risk model to obtain a group insurance claim model; and inputting the target sample of the target group insurance customer into the group insurance claim model to obtain a group insurance claim risk prediction result. According to the invention, the accuracy of group insurance claim risk assessment is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, device, computer equipment and storage medium for processing a claims risk model. Background Art

[0002] As the demand for insurance continues to increase, the number of insurance claims is also growing. In the insurance industry, modeling and claim risk prediction for customers is conducive to customer risk management, reducing the number of insurance claims and reducing the losses of insurance companies.

[0003] The customers of group insurance are usually B-side customers (i.e. corporate customers, including various companies, enterprises, institutions and other organizations, rather than individual consumers). The number of group insurance policies is smaller than that of personal insurance, but the diversity of insurance types is higher and the demand for customization is higher. This also leads to high dimensionality and high noise in the data when group insurance claims are settled. There are a large number of unique features in the data that cannot be reused, which brings great challenges to model training, and there are usually many outliers in group insurance data. Traditional machine learning methods are often difficult to process group insurance data and have low accuracy. Summary of the invention

[0004] The purpose of the embodiments of the present application is to propose a claims risk model processing method, apparatus, computer equipment and storage medium to solve the problem of low accuracy in group insurance claims risk assessment.

[0005] In order to solve the above technical problems, the present application embodiment provides a method for processing a claims risk model, which adopts the following technical solution:

[0006] Obtaining a first training set, where the first training set includes original samples and risk labels of multiple group insurance customers;

[0007] Training a first initial risk model according to the first training set to obtain a first risk model, and obtaining a risk prediction result of the first risk model for each group insurance customer;

[0008] Constructing a second training set according to the obtained risk prediction result and the first training set, and training a second initial risk model according to the second training set to obtain a second risk model;

[0009] Based on the integrated algorithm and the second training set, the first risk model and the second risk model are integratedly trained to obtain an initial group insurance claims model;

[0010] Cross-validating the first risk model and the second risk model in the initial group insurance claim settlement model, and adjusting the model weights of the first risk model and the second risk model according to the obtained cross-validation results to obtain a group insurance claim settlement model;

[0011] A target sample of target group insurance customers is obtained, and the target sample is input into the group insurance claim settlement model to obtain a group insurance claim settlement risk prediction result.

[0012] In order to solve the above technical problems, the embodiment of the present application also provides a claim risk model processing device, which adopts the following technical solution:

[0013] A training set acquisition module, used to acquire a first training set, wherein the first training set includes original samples and risk labels of multiple group insurance customers;

[0014] A first training module is used to train a first initial risk model according to the first training set to obtain a first risk model, and obtain the risk prediction result of the first risk model for each group insurance customer;

[0015] A second training module, used for constructing a second training set according to the obtained risk prediction result and the first training set, and training a second initial risk model according to the second training set to obtain a second risk model;

[0016] An integrated training module, configured to perform integrated training on the first risk model and the second risk model based on an integrated algorithm and the second training set to obtain an initial group insurance claims model;

[0017] A cross-validation module, used to cross-validate the first risk model and the second risk model in the initial group insurance claim settlement model, and adjust the model weights of the first risk model and the second risk model according to the obtained cross-validation results to obtain a group insurance claim settlement model;

[0018] The risk prediction module is used to obtain a target sample of target group insurance customers and input the target sample into the group insurance claim model to obtain a group insurance claim risk prediction result.

[0019] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0020] Obtaining a first training set, where the first training set includes original samples and risk labels of multiple group insurance customers;

[0021] Training a first initial risk model according to the first training set to obtain a first risk model, and obtaining a risk prediction result of each group insurance customer by the first risk model;

[0022] Constructing a second training set according to the obtained risk prediction result and the first training set, and training a second initial risk model according to the second training set to obtain a second risk model;

[0023] Based on the integrated algorithm and the second training set, the first risk model and the second risk model are integratedly trained to obtain an initial group insurance claims model;

[0024] Cross-validating the first risk model and the second risk model in the initial group insurance claim settlement model, and adjusting the model weights of the first risk model and the second risk model according to the obtained cross-validation results to obtain a group insurance claim settlement model;

[0025] A target sample of target group insurance customers is obtained, and the target sample is input into the group insurance claim settlement model to obtain a group insurance claim settlement risk prediction result.

[0026] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0027] Obtaining a first training set, where the first training set includes original samples and risk labels of multiple group insurance customers;

[0028] Training a first initial risk model according to the first training set to obtain a first risk model, and obtaining a risk prediction result of the first risk model for each group insurance customer;

[0029] Constructing a second training set according to the obtained risk prediction result and the first training set, and training a second initial risk model according to the second training set to obtain a second risk model;

[0030] Based on the integrated algorithm and the second training set, the first risk model and the second risk model are integratedly trained to obtain an initial group insurance claims model;

[0031] Cross-validating the first risk model and the second risk model in the initial group insurance claim settlement model, and adjusting the model weights of the first risk model and the second risk model according to the obtained cross-validation results to obtain a group insurance claim settlement model;

[0032] A target sample of target group insurance customers is obtained, and the target sample is input into the group insurance claim settlement model to obtain a group insurance claim settlement risk prediction result.

[0033] Compared with the prior art, the embodiments of the present application mainly have the following beneficial effects: obtaining original samples containing multiple group insurance customers and risk labels, the risk labels indicating whether the group insurance customers have ever made claims, and if so, indicating that there is a risk; training a first initial risk model according to a first training set to obtain a first risk model, and obtaining the risk prediction results of the first risk model for each group insurance customer; constructing a second training set according to the risk prediction results and the first training set to provide additional features for subsequent models; training a second initial risk model according to the second training set to obtain a second risk model; based on the integrated algorithm and the second training set, integrating the first risk model and the second risk model, combining different models to enhance the model The overall prediction ability is used to obtain the initial group insurance claims model; the first risk model and the second risk model in the initial group insurance claims model are cross-validated, and the model weights of the first risk model and the second risk model are adjusted according to the cross-validation results to further optimize the model performance and obtain the group insurance claims model; the target sample of target group insurance customers is obtained and input into the group insurance claims model to obtain the group insurance claims risk prediction results; through training set construction, multi-model integration and cross-validation weight adjustment, it can effectively deal with data problems with high dimensions, high noise and personalized characteristics, meet the insurance industry's demand for accurate risk assessment, improve the accuracy and robustness of group insurance claims risk prediction, and reduce the claims risk of insurance companies. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0035] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0036] Figure 2 is a flow chart of an embodiment of a method for processing a claims risk model according to the present application;

[0037] Figure 3 is a schematic diagram of the structure of an embodiment of a claim risk model processing device according to the present application;

[0038] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0039] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0040] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0041] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0042] like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables.

[0043] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0044] The terminal devices 101, 102, and 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, etc.

[0045] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .

[0046] It should be noted that the claim risk model processing method provided in the embodiment of the present application is generally executed by a server, and accordingly, the claim risk model processing device is generally set in the server.

[0047] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0048] Continue to refer Figure 2 , shows a flow chart of an embodiment of a method for processing a claim risk model according to the present application. The method for processing a claim risk model comprises the following steps:

[0049] Step S201, obtaining a first training set, the first training set including original samples and risk labels of multiple group insurance customers.

[0050] In this embodiment, the claim risk model processing method is executed on an electronic device (eg Figure 1 The server shown in the figure) can communicate with the terminal device through a wired connection or a wireless connection. It should be noted that the above wireless connection method may include but is not limited to 3G / 4G / 5G connection, Wi-Fi connection, Bluetooth connection, Wi MAX connection, Zigbee connection, UWB (Ultra Wi-Fi Band) connection, and other wireless connection methods currently known or to be developed in the future.

[0051] Specifically, first, a first training set containing multiple original samples of group insurance customers and corresponding risk labels is obtained. The original samples include the basic information of the customers, historical group insurance data and risk labels. The risk label indicates whether the insurance company has ever paid compensation to group insurance customers in the past. If so, the group insurance customer is a high-risk customer.

[0052] Step S202: training a first initial risk model according to a first training set to obtain a first risk model, and obtaining a risk prediction result of the first risk model for each group insurance customer.

[0053] Specifically, the first initial risk model is trained using the first training set, and the first risk model is obtained after the training is completed. Through training, the first risk model can learn the relationship between each feature of each original sample in the first training set and the risk label. Subsequently, the trained first risk model is used to predict the risk of each group insurance customer to obtain a preliminary risk prediction result.

[0054] Step S203, constructing a second training set according to the obtained risk prediction result and the first training set, and training a second initial risk model according to the second training set to obtain a second risk model.

[0055] Specifically, the risk prediction results obtained from the first risk model are combined with the original first training set to obtain a second training set, which contains not only the original feature data but also the prediction results provided by the first risk model. The second initial risk model is then trained using the second training set to obtain a second risk model.

[0056] In one embodiment, the first initial risk model / first risk model may be a LGBM (Light Grid Boosting Machine) model. The LGBM model is an efficient machine learning algorithm based on a gradient boosting framework, which is suitable for processing large-scale data sets and high-dimensional features. The LGBM model can be used for data classification, regression prediction and other tasks. The second initial risk model / second risk model may be a logistic regression model. The logistic regression model is a generalized linear model used to solve binary classification problems. It performs classification by mapping the output of the linear regression model to a probability value, and uses the sigmoid function to convert the output of the linear regression into a probability value between 0 and 1, thereby indicating the probability that the sample belongs to a certain class. In the field of finance and insurance, the customer object can be analyzed by a logistic regression model to predict whether the customer object has a default or payment risk, thereby helping financial institutions make risk assessments and decisions. It can be understood that the first risk model may also be a logistic regression model, and the second risk model may also be a LGBM model; the first risk model and the second risk model may also be other types of models, which are not limited in this application.

[0057] Furthermore, the above-mentioned step of constructing a second training set based on the obtained risk prediction results and the first training set may include: for each group insurance customer in the first training set, obtaining the risk prediction result of the group insurance customer; adding the risk prediction result as a sample feature to the original sample of the group insurance customer to obtain a customer sample of the group insurance customer; and generating a second training set based on the customer samples of each group insurance customer.

[0058] Specifically, for each group insurance customer in the first training set, a risk prediction result thereof is obtained through the first risk model. This result is a risk assessment of the group insurance customer by the first risk model, indicating the risk level of a possible claim by the group insurance customer.

[0059] The risk prediction results are added as new feature values ​​to the original sample of group insurance customers. The new features provide additional information that can help the second risk model better capture the risk characteristics of group insurance customers. Based on the new customer samples, a richer second training set can be constructed; the second training set contains the original samples and the newly added risk prediction results, providing a more comprehensive information basis for the subsequent second initial risk model training, so that the second risk model can use the information in the first risk model to improve its ability to predict risks.

[0060] In this embodiment, the risk prediction results of the first risk model are added to the original sample as new features to generate a second training set containing more information, which is conducive to improving the prediction ability and accuracy of the second risk model and optimizing the model performance.

[0061] Step S204: Based on the integrated algorithm and the second training set, the first risk model and the second risk model are integratedly trained to obtain an initial group insurance claims model.

[0062] Among them, the ensemble algorithm is a machine learning strategy that completes the learning task by building and combining multiple base evaluators. The core idea of ​​the ensemble algorithm is to achieve better prediction performance than a single "weak classifier" by combining the prediction results of multiple "weak classifiers", thereby forming a "strong classifier".

[0063] Specifically, based on the integrated algorithm and the second training set, the first risk model and the second risk model are integrated and trained. Integrated training can combine the advantages of each model to create an integrated model (initial group insurance claims model) with better performance. The integrated algorithm can effectively adjust the contribution of each model so that the overall prediction ability of the integrated model exceeds that of a single model.

[0064] In one embodiment, the integrated algorithm may use the AdaBoost algorithm in the Boosting algorithm.

[0065] Step S205, cross-validating the first risk model and the second risk model in the initial group insurance claim settlement model, and adjusting the model weights of the first risk model and the second risk model according to the obtained cross-validation results to obtain the group insurance claim settlement model.

[0066] Specifically, the initial group insurance claims model is cross-validated, the second training set is divided into several subsets, the first risk model and the second risk model in the initial group insurance claims model are trained and validated on different subsets, and the cross-validation results are obtained.

[0067] According to the cross-validation results, the performance of the first risk model and the second risk model in the initial group insurance claims model can be evaluated, and the model weights of the first risk model and the second risk model can be adjusted according to the performance to obtain the group insurance claims model. The purpose of adjusting the weights is to optimize the contribution of each model in the initial group insurance claims model and further improve the prediction performance of the initial group insurance claims model.

[0068] Step S206, obtaining a target sample of target group insurance customers, and inputting the target sample into a group insurance claim model to obtain a group insurance claim risk prediction result.

[0069] Specifically, in practical applications, target samples of target group insurance customers are obtained and input into the final optimized group insurance claims model. Through the prediction of the model, the group insurance claims risk prediction results of the target group insurance customers are obtained. The group insurance claims risk prediction results indicate the risk of claims for the target group insurance customers, which can provide risk assessment for insurance companies, help them conduct risk management and decision-making, and reduce claims risks and losses.

[0070] In this embodiment, an original sample including multiple group insurance customers and a risk label are obtained, and the risk label indicates whether the group insurance customer has ever made a claim, and if so, it indicates that there is a risk; a first initial risk model is trained according to the first training set to obtain a first risk model, and the risk prediction result of the first risk model for each group insurance customer is obtained; a second training set is constructed according to the risk prediction result and the first training set to provide additional features for subsequent models; a second initial risk model is trained according to the second training set to obtain a second risk model; based on the integrated algorithm and the second training set, the first risk model and the second risk model are integratedly trained to combine different models to enhance the overall prediction ability of the model, and obtain Initial group insurance claims model; cross-validate the first risk model and the second risk model in the initial group insurance claims model, and adjust the model weights of the first risk model and the second risk model according to the cross-validation results, further optimize the model performance and obtain the group insurance claims model; obtain the target sample of target group insurance customers and input it into the group insurance claims model to obtain the group insurance claims risk prediction results; through training set construction, multi-model integration and cross-validation weight adjustment, it can effectively deal with data problems with high dimensions, high noise and personalized characteristics, meet the insurance industry's needs for accurate risk assessment, improve the accuracy and robustness of group insurance claims risk prediction, and reduce the claims risk of insurance companies.

[0071] Furthermore, the above-mentioned step of training the second initial risk model according to the second training set to obtain the second risk model may include: parsing the second training set to obtain training influence parameters of the second training set, the training influence parameters including the sample data volume and the positive and negative sample ratio of the second training set; determining the training configuration information of the second initial risk model according to the training influence parameters, the training configuration information including model configuration information determined according to the sample data volume, and category weights of customer samples of each group insurance customer determined according to the positive and negative sample ratio; training the second initial risk model based on the training configuration information and the second training set to obtain the second risk model.

[0072] Specifically, the second training set is parsed to obtain training influencing parameters. The training influencing parameters include sample data volume and positive and negative sample ratio; sample data volume is the amount of data included in the second training set, such as the number of samples and the total amount of data; the positive and negative sample ratio represents the ratio between samples with risk labels of positive (high risk) and negative (low risk).

[0073] The training configuration information of the second initial risk model is determined according to the training influencing parameters. The training configuration information includes model configuration information and category weights. Among them, the model configuration information is determined according to the amount of sample data, which refers to the setting information of the second initial risk model, including learning rate, optimization algorithm, regularization type and maximum number of iterations. For example, before training, the learning rate of the second initial risk model can be set to 0.01; the sag (Stochast ic Average Gradient) optimization algorithm is used to apply to large-scale data sets; the regularization type uses L2 regression to prevent overfitting. L2 regularization penalizes excessive coefficients by adding the square sum of parameters to the loss function, thereby making the model simpler and smoother; the maximum number of iterations of the model max_i ter is set to the default 10. If the model has not converged after 10 iterations, the iteration is stopped.

[0074] The class weight is determined based on the ratio of positive and negative samples. Each customer sample has a class weight, which is used to balance the imbalance of positive and negative samples during training, so that the model can predict the minority class samples more accurately. For example, if the ratio of positive and negative samples is positive sample: negative sample = 1:9, then the class weight of the customer sample with a risk label of 0 (paid case) is set to 0.9, and the class weight of the customer sample with a risk label of 1 (unpaid case) is set to 0.1, which means that the weight of the paid case is 9 times that of the unpaid case, thereby giving more attention to the fewer paid cases during model training and avoiding the model from being biased towards the majority class.

[0075] The second initial risk model is trained using the determined training configuration information and the second training set to obtain a second risk model.

[0076] In this embodiment, the second training set is parsed to obtain training influencing parameters, including the sample data volume and the positive and negative sample ratio of the second training set; training configuration information is determined based on the training influencing parameters, including model configuration information determined based on the sample data volume, and category weights of customer samples of each group insurance customer determined based on the positive and negative sample ratio; the training configuration information optimizes the training process of the second initial risk model, improves the model's adaptability to different samples and prediction accuracy, and ensures the accuracy of subsequent claims risk prediction.

[0077] Furthermore, the above-mentioned step S204 may include: when conducting the first round of training, determining the initial sample weight corresponding to each customer sample in the second training set; training the first model according to each customer sample with the initial sample weight, and obtaining the first prediction result of the first model for each customer sample, wherein the first model is a first risk model or a second risk model; calculating the error rate of the first model according to the first prediction result and its risk label of each customer sample, and calculating the weak classifier weight of the first model according to the error rate; calculating the updated sample weight of each customer sample according to the weak classifier weight, the initial sample weight of each customer sample, the first prediction result and its risk label; when conducting each round of training after the first round of training, using the updated sample weight of each customer sample as the new initial sample weight, and training the first risk model and the second risk model alternately to obtain the initial group insurance claims model.

[0078] Specifically, in the first round of ensemble training, an initial sample weight is determined for each customer sample in the second training set. Usually, the initial sample weight of each customer sample in the first round of training is equal, usually set to Where N is the total number of customer samples, which ensures that each customer sample has the same impact on model training in the initial stage.

[0079] The first model is trained according to each customer sample with the initial sample weight, and after the training is completed, the first prediction result of the first model for each customer sample is obtained. The first model here can be a first risk model or a second risk model. That is, in the first round of ensemble training, the first risk model can be trained first, or the second risk model can be trained first.

[0080] Then, the error rate of the first model is calculated based on the first prediction results of each customer sample and its risk label. The error rate reflects the accuracy of the model prediction results. In one embodiment, the error rate is defined as the sum of the weights of samples misclassified by the weak classifier (the first risk model or the second risk model):

[0081]

[0082] Among them, ε represents the error rate, ω i is the initial sample weight of customer sample i, N is the total number of customer samples, I(y i ≠h(x i )) is an indicator function, which is 1 when customer sample i is misclassified and 0 otherwise.

[0083] The weak classifier weight of the first model is calculated based on the error rate, and the weak classifier weight reflects the importance of the model in the integration. In one embodiment, the weak classifier weight calculation formula is as follows:

[0084]

[0085] Among them, α represents the weak classifier weight of the first model, and ε represents the error rate of the first model.

[0086] Then, based on the weak classifier weights, the initial sample weights corresponding to each customer sample, the first prediction results, and the risk labels, the updated sample weights of each customer sample are calculated. The updated sample weights are used for the next round of training to adjust the influence of customer samples in training. When updating the sample weights, the sample weights of misclassified samples are strengthened and the sample weights of correctly classified samples are weakened, so that more attention is paid to misclassified customer samples in the next round of training, while trying to avoid abnormal samples from getting higher weights.

[0087] Based on the first prediction result and the risk label, it can be determined whether the customer sample is classified correctly; the customer samples that are correctly classified are:

[0088] ω i+1 =ω i *exp(-α)

[0089] Among them, ω i+1 represents the updated sample weight of customer sample i, ω i represents the initial sample weight of customer sample i, and α represents the weak classifier weight of the first model.

[0090] The samples of misclassified customers are:

[0091] ω i+1 =ω i *exp(α)

[0092] Among them, ω i+1 represents the updated sample weight of customer sample i, ω i represents the initial sample weight of customer sample i, and α represents the weak classifier weight of the first model.

[0093] In each round of training after the first round of training, the updated sample weights obtained in the previous round are used as new initial sample weights to train the first risk model and the second risk model alternately. That is, in each round of training, only one of the first risk model and the second risk model is trained; and if the first risk model (or the second risk model) is trained in the X+1th round, the second risk model (or the first risk model) must be trained in the Xth and X+2th rounds. This process is repeated until the initial group insurance claims model is obtained.

[0094] In this embodiment, by setting the initial sample weights, calculating the error rate and updating the sample weights, the performance of the first risk model or the second risk model in each round of training is optimized, the prediction accuracy of the overall model is improved, the adaptability of the overall model to different samples is enhanced, and the effect of subsequent claims risk prediction is ensured.

[0095] Furthermore, the above-mentioned step of cross-validating the first risk model and the second risk model in the initial group insurance claims model may include: dividing the second training set into several sub-training sets; for each sub-training set, constructing the sub-training set as a cross-validation set, and constructing the remaining sub-training sets as cross-training sets; training the first risk model and the second risk model in the initial group insurance claims model according to the cross-training set, and obtaining the cross-validation results of the first risk model and the second risk model on the cross-validation set after the training is completed.

[0096] Specifically, during cross-validation, the second training set is first divided into several sub-training sets (for example, 5 or 10). For each sub-training set, it can be constructed as a cross-validation set, and the remaining sub-training sets can be constructed as cross-training sets. The cross-training set is used to train the model, and the cross-validation set is used to evaluate the model performance after training. Assuming there are K sub-training sets, a total of K cross-training sets and their corresponding cross-validation sets are now constructed.

[0097] For each cross-training set, the first risk model and the second risk model in the initial group insurance claims model are trained according to the cross-training set. After the training is completed, the cross-validation results of the first risk model and the second risk model on the cross-validation set are obtained, that is, the prediction results of the first risk model and the second risk model on whether there is a claim risk for each customer sample in the cross-validation set.

[0098] In one embodiment, the second training set is divided into K sub-training sets, and the cross-circular validation is performed for K training and validation. In the i-th iteration (i=1, 2, ..., K), the i-th sub-training set is selected as the cross-validation set, and the remaining K-1 sub-training sets are combined into the cross-training set; then the first risk model and the second risk model are trained on the cross-training set, and the first risk model and the second risk model are used to perform prediction on the cross-validation set.

[0099] In this embodiment, during cross-validation, the second training set is divided into multiple sub-training sets, and the model is repeatedly trained and verified on different sub-training sets to ensure the generalization ability and stability of the model, thereby improving the accuracy and robustness of claims risk prediction.

[0100] Furthermore, the above-mentioned step of adjusting the model weights of the first risk model and the second risk model according to the obtained cross-validation results to obtain the group insurance claims model may include: respectively obtaining the cross-validation results of the first risk model and the second risk model on each cross-validation set; respectively calculating the average performance of the first risk model and the second risk model in the cross-validation according to the obtained cross-validation results; determining the model weights of the first risk model and the second risk model according to the obtained average performance, and generating the group insurance claims model according to the determined model weights.

[0101] Specifically, the cross-validation results of the first risk model and the second risk model on each cross-validation set are obtained respectively. According to the cross-validation results, the prediction performance indicators of the first risk model / the second risk model on different cross-validation sets, such as accuracy, recall, etc., can be determined. Therefore, based on the obtained cross-validation results, the average performance of the first risk model and the second risk model in the cross-validation can be calculated respectively. The average performance represents the overall performance of the first risk model / the second risk model on each cross-validation set, and is a key indicator for evaluating the stability and generalization ability of the model.

[0102] According to the obtained average performance, the model weights of the first risk model and the second risk model in the integrated model can be determined. The model weights reflect the contribution of the first risk model and the second risk model in the final prediction. The higher the performance of the model, the greater the weight, and vice versa.

[0103] The average performance can be a numeric value. The direct proportional method can be used to determine the model weights: weights are assigned according to the ratio of the average performance; for example, if the average performance of the first risk model is 0.8 and the average performance of the second risk model is 0.7, the model weight ratio is 0.8:0.7. It can be standardized to 0.53:0.47. The inverse error method can also be used: weights are assigned according to the error rate of the model (1-average performance), and models with smaller errors have higher weights.

[0104] After determining the model weights of the first risk model and the second risk model in the integrated model, the final group insurance claims model can be generated. The group insurance claims model combines the advantages of the first risk model and the second risk model to ensure the accuracy and reliability of claims risk prediction.

[0105] In this embodiment, the model weights of the first risk model and the second risk model are adjusted according to the cross-validation results, and the combination of risk models is optimized, so that the final group insurance claim settlement model has higher prediction accuracy and stability.

[0106] Furthermore, the above-mentioned step of inputting the target sample into the group insurance claim model to obtain the group insurance claim risk prediction result may include: inputting the target sample into the first risk model in the group insurance claim model to obtain the second prediction result; inputting the target sample into the second risk model in the group insurance claim model to obtain the third prediction result; and performing weighted calculation on the second prediction result and the third prediction result according to the model weights of the first risk model and the second risk model to obtain the group claim risk prediction result.

[0107] Specifically, the target sample is input into the first risk model in the group insurance claim model for risk prediction to obtain the second prediction result. The target sample is input into the second risk model in the group insurance claim model for risk prediction to obtain the third prediction result. According to the model weights of the first risk model and the second risk model, the second prediction result and the third prediction result are weighted and calculated to obtain the final group claim risk prediction result. The weighted average can effectively utilize the advantages of the first risk model and the second risk model to improve the overall prediction performance.

[0108] In this embodiment, the target samples are input into two risk models respectively, and the final prediction results are weighted and calculated according to the model weights, which comprehensively utilizes the advantages of the two risk models and improves the accuracy of group insurance claim risk prediction.

[0109] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned first training set, the above-mentioned first training set can also be stored in a node of a blockchain.

[0110] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0111] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0112] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0113] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0114] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by 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 they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0115] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a claim risk model processing device, and the device embodiment is similar to Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0116] like Figure 3 As shown, the claim risk model processing device 300 described in this embodiment includes: a training set acquisition module 301, a first training module 302, a second training module 303, an integrated training module 304, a cross-validation module 305 and a risk prediction module 306, wherein:

[0117] The training set acquisition module 301 is used to acquire a first training set, which includes original samples and risk labels of multiple group insurance customers.

[0118] The first training module 302 is used to train a first initial risk model according to a first training set to obtain a first risk model, and obtain the risk prediction result of the first risk model for each group insurance customer.

[0119] The second training module 303 is used to construct a second training set according to the obtained risk prediction result and the first training set, and train a second initial risk model according to the second training set to obtain a second risk model.

[0120] The integrated training module 304 is used to perform integrated training on the first risk model and the second risk model based on the integrated algorithm and the second training set to obtain an initial group insurance claim model.

[0121] The cross-validation module 305 is used to cross-validate the first risk model and the second risk model in the initial group insurance claim settlement model, and adjust the model weights of the first risk model and the second risk model according to the obtained cross-validation results to obtain the group insurance claim settlement model.

[0122] The risk prediction module 306 is used to obtain a target sample of target group insurance customers and input the target sample into a group insurance claim model to obtain a group insurance claim risk prediction result.

[0123] In this embodiment, an original sample including multiple group insurance customers and a risk label are obtained, and the risk label indicates whether the group insurance customer has ever made a claim, and if so, it indicates that there is a risk; a first initial risk model is trained according to the first training set to obtain a first risk model, and the risk prediction result of the first risk model for each group insurance customer is obtained; a second training set is constructed according to the risk prediction result and the first training set to provide additional features for subsequent models; a second initial risk model is trained according to the second training set to obtain a second risk model; based on the integrated algorithm and the second training set, the first risk model and the second risk model are integratedly trained to combine different models to enhance the overall prediction ability of the model, and obtain Initial group insurance claims model; cross-validate the first risk model and the second risk model in the initial group insurance claims model, and adjust the model weights of the first risk model and the second risk model according to the cross-validation results, further optimize the model performance and obtain the group insurance claims model; obtain the target sample of target group insurance customers and input it into the group insurance claims model to obtain the group insurance claims risk prediction results; through training set construction, multi-model integration and cross-validation weight adjustment, it can effectively deal with data problems with high dimensions, high noise and personalized characteristics, meet the insurance industry's needs for accurate risk assessment, improve the accuracy and robustness of group insurance claims risk prediction, and reduce the claims risk of insurance companies.

[0124] In some optional implementations of this embodiment, the second training module 303 may include: a prediction acquisition submodule, a prediction addition submodule, and a training set generation submodule, wherein:

[0125] The prediction acquisition submodule is used to obtain the risk prediction result of the group insurance customer for each group insurance customer in the first training set.

[0126] The prediction adding submodule is used to add the risk prediction results as sample features to the original sample of group insurance customers to obtain the customer sample of group insurance customers.

[0127] The training set generation submodule is used to generate a second training set based on customer samples of each group insurance customer.

[0128] In this embodiment, the risk prediction results of the first risk model are added to the original sample as new features to generate a second training set containing more information, which is conducive to improving the prediction ability and accuracy of the second risk model and optimizing the model performance.

[0129] In some optional implementations of this embodiment, the second training module 303 may include: a training set parsing submodule, a configuration determination submodule and a second training submodule, wherein:

[0130] The training set parsing submodule is used to parse the second training set to obtain the training influencing parameters of the second training set, where the training influencing parameters include the sample data volume and the positive and negative sample ratio of the second training set.

[0131] The configuration determination submodule is used to determine the training configuration information of the second initial risk model according to the training influencing parameters. The training configuration information includes model configuration information determined according to the sample data volume and category weights of customer samples of each group insurance customer determined according to the ratio of positive and negative samples.

[0132] The second training submodule is used to train a second initial risk model based on the training configuration information and the second training set to obtain a second risk model.

[0133] In this embodiment, the second training set is parsed to obtain training influencing parameters, including the sample data volume and the positive and negative sample ratio of the second training set; training configuration information is determined based on the training influencing parameters, including model configuration information determined based on the sample data volume, and category weights of customer samples of each group insurance customer determined based on the positive and negative sample ratio; the training configuration information optimizes the training process of the second initial risk model, improves the model's adaptability to different samples and prediction accuracy, and ensures the accuracy of subsequent claims risk prediction.

[0134] In some optional implementations of this embodiment, the integrated training module 304 may include: a weight determination submodule, a model training submodule, a weight calculation submodule, a weight update submodule, and a rotation training submodule, wherein:

[0135] The weight determination submodule is used to determine the initial sample weight corresponding to each customer sample in the second training set during the first round of training.

[0136] The model training submodule is used to train the first model according to each customer sample with an initial sample weight, and obtain the first prediction result of the first model for each customer sample, wherein the first model is a first risk model or a second risk model.

[0137] The weight calculation submodule is used to calculate the error rate of the first model according to the first prediction result of each customer sample and its risk label, and calculate the weak classifier weight of the first model according to the error rate.

[0138] The weight updating submodule is used to calculate the updated sample weight of each customer sample according to the weak classifier weight, the initial sample weight of each customer sample, the first prediction result and its risk label.

[0139] The rotation training submodule is used to use the updated sample weights of each customer sample as new initial sample weights in each round of training after the first round of training, and to train the first risk model and the second risk model in turn to obtain an initial group insurance claims model.

[0140] In this embodiment, by setting the initial sample weights, calculating the error rate and updating the sample weights, the performance of the first risk model or the second risk model in each round of training is optimized, the prediction accuracy of the overall model is improved, the adaptability of the overall model to different samples is enhanced, and the effect of subsequent claims risk prediction is ensured.

[0141] In some optional implementations of this embodiment, the cross-validation module 305 may include: a training set partitioning submodule, a subset construction submodule, and a cross-training submodule, wherein:

[0142] The training set division submodule is used to divide the second training set into a plurality of sub-training sets.

[0143] The subset construction submodule is used to construct the sub-training set as a cross-validation set for each sub-training set, and to construct the remaining sub-training sets as cross-training sets.

[0144] The cross-training submodule is used to train the first risk model and the second risk model in the initial group insurance claims model according to the cross-training set, and obtain the cross-validation results of the first risk model and the second risk model on the cross-validation set after the training.

[0145] In this embodiment, during cross-validation, the second training set is divided into multiple sub-training sets, and the model is repeatedly trained and verified on different sub-training sets to ensure the generalization ability and stability of the model, thereby improving the accuracy and robustness of claims risk prediction.

[0146] In some optional implementations of this embodiment, the cross-validation module 305 may include: a result acquisition submodule, a performance calculation submodule, and a weight determination submodule, wherein:

[0147] The result acquisition submodule is used to respectively obtain the cross-validation results of the first risk model and the second risk model on each cross-validation set.

[0148] The performance calculation submodule is used to calculate the average performance of the first risk model and the second risk model in the cross validation according to the obtained cross validation results.

[0149] The weight determination submodule is used to determine the model weights of the first risk model and the second risk model according to the obtained average performance, and to generate a group insurance claim model according to the determined model weights.

[0150] In this embodiment, the model weights of the first risk model and the second risk model are adjusted according to the cross-validation results, and the combination of risk models is optimized, so that the final group insurance claim settlement model has higher prediction accuracy and stability.

[0151] In some optional implementations of this embodiment, the risk prediction module 306 may include: a first prediction submodule, a second prediction submodule, and a result generation submodule, wherein:

[0152] The first prediction submodule is used to input the target sample into the first risk model in the group insurance claim model to obtain a second prediction result.

[0153] The second prediction submodule is used to input the target sample into the second risk model in the group insurance claim model to obtain a third prediction result.

[0154] The result generation submodule is used to perform weighted calculation on the second prediction result and the third prediction result according to the model weights of the first risk model and the second risk model to obtain a group claim risk prediction result.

[0155] In this embodiment, the target samples are input into two risk models respectively, and the final prediction results are weighted and calculated according to the model weights, which comprehensively utilizes the advantages of the two risk models and improves the accuracy of group insurance claim risk prediction.

[0156] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0157] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 having components 41-43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital processors (DSPs), embedded devices, etc.

[0158] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0159] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the computer device 4. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of the claim risk model processing method, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0160] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run computer-readable instructions stored in the memory 41 or process data, such as computer-readable instructions for running the claim risk model processing method.

[0161] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0162] The computer device provided in this embodiment can execute the above-mentioned claim risk model processing method. This claim risk model processing method can be the claim risk model processing method of each of the above-mentioned embodiments.

[0163] In this embodiment, an original sample including multiple group insurance customers and a risk label are obtained, and the risk label indicates whether the group insurance customer has ever made a claim, and if so, it indicates that there is a risk; a first initial risk model is trained according to the first training set to obtain a first risk model, and the risk prediction result of the first risk model for each group insurance customer is obtained; a second training set is constructed according to the risk prediction result and the first training set to provide additional features for subsequent models; a second initial risk model is trained according to the second training set to obtain a second risk model; based on the integrated algorithm and the second training set, the first risk model and the second risk model are integratedly trained to combine different models to enhance the overall prediction ability of the model, and obtain Initial group insurance claims model; cross-validate the first risk model and the second risk model in the initial group insurance claims model, and adjust the model weights of the first risk model and the second risk model according to the cross-validation results, further optimize the model performance and obtain the group insurance claims model; obtain the target sample of target group insurance customers and input it into the group insurance claims model to obtain the group insurance claims risk prediction results; through training set construction, multi-model integration and cross-validation weight adjustment, it can effectively deal with data problems with high dimensions, high noise and personalized characteristics, meet the insurance industry's needs for accurate risk assessment, improve the accuracy and robustness of group insurance claims risk prediction, and reduce the claims risk of insurance companies.

[0164] The present application also provides another embodiment, namely, providing a computer-readable storage medium, which stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the claim risk model processing method as described above.

[0165] In this embodiment, an original sample including multiple group insurance customers and a risk label are obtained, and the risk label indicates whether the group insurance customer has ever made a claim, and if so, it indicates that there is a risk; a first initial risk model is trained according to the first training set to obtain a first risk model, and the risk prediction result of the first risk model for each group insurance customer is obtained; a second training set is constructed according to the risk prediction result and the first training set to provide additional features for subsequent models; a second initial risk model is trained according to the second training set to obtain a second risk model; based on the integrated algorithm and the second training set, the first risk model and the second risk model are integratedly trained to combine different models to enhance the overall prediction ability of the model, and obtain Initial group insurance claims model; cross-validate the first risk model and the second risk model in the initial group insurance claims model, and adjust the model weights of the first risk model and the second risk model according to the cross-validation results, further optimize the model performance and obtain the group insurance claims model; obtain the target sample of target group insurance customers and input it into the group insurance claims model to obtain the group insurance claims risk prediction results; through training set construction, multi-model integration and cross-validation weight adjustment, it can effectively deal with data problems with high dimensions, high noise and personalized characteristics, meet the insurance industry's needs for accurate risk assessment, improve the accuracy and robustness of group insurance claims risk prediction, and reduce the claims risk of insurance companies.

[0166] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0167] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A method for processing a claims risk model, characterized in that: The steps include: Obtaining a first training set, where the first training set includes original samples and risk labels of multiple group insurance customers; Training a first initial risk model according to the first training set to obtain a first risk model, and obtaining a risk prediction result of the first risk model for each group insurance customer; Constructing a second training set according to the obtained risk prediction result and the first training set, and training a second initial risk model according to the second training set to obtain a second risk model; Based on the integrated algorithm and the second training set, the first risk model and the second risk model are integratedly trained to obtain an initial group insurance claims model; Cross-validating the first risk model and the second risk model in the initial group insurance claim settlement model, and adjusting the model weights of the first risk model and the second risk model according to the obtained cross-validation results to obtain a group insurance claim settlement model; A target sample of target group insurance customers is obtained, and the target sample is input into the group insurance claim settlement model to obtain a group insurance claim settlement risk prediction result.

2. The claim risk model processing method according to claim 1, characterized in that: The step of constructing a second training set according to the obtained risk prediction result and the first training set comprises: For each group insurance customer in the first training set, obtaining a risk prediction result of the group insurance customer; Adding the risk prediction results as sample features to the original sample of the group insurance customers to obtain a customer sample of the group insurance customers; A second training set is generated based on customer samples of each group insurance customer.

3. The claim risk model processing method according to claim 1, characterized in that: The step of training a second initial risk model according to the second training set to obtain a second risk model comprises: Parsing the second training set to obtain training influencing parameters of the second training set, wherein the training influencing parameters include the sample data volume and the positive and negative sample ratio of the second training set; Determining training configuration information of a second initial risk model according to the training influencing parameters, the training configuration information including model configuration information determined according to the sample data volume, and category weights of customer samples of each group insurance customer determined according to the positive and negative sample ratios; The second initial risk model is trained based on the training configuration information and the second training set to obtain a second risk model.

4. The claim risk model processing method according to claim 1, characterized in that: The step of performing integrated training on the first risk model and the second risk model based on the integrated algorithm and the second training set to obtain an initial group insurance claims model comprises: During the first round of training, determining the initial sample weights corresponding to each customer sample in the second training set; Training a first model according to each customer sample with an initial sample weight, and obtaining a first prediction result of the first model for each customer sample, wherein the first model is the first risk model or the second risk model; Calculating the error rate of the first model according to the first prediction results of each customer sample and the risk label thereof, and calculating the weak classifier weight of the first model according to the error rate; Calculate the updated sample weight of each customer sample according to the weak classifier weight, the initial sample weight of each customer sample, the first prediction result and the risk label thereof; In each round of training after the first round of training, the updated sample weights of each customer sample are used as new initial sample weights, and the first risk model and the second risk model are trained alternately to obtain an initial group insurance claims model.

5. The claim risk model processing method according to claim 1, characterized in that: The step of cross-validating the first risk model and the second risk model in the initial group insurance claims model comprises: Dividing the second training set into a plurality of sub-training sets; For each sub-training set, construct the sub-training set as a cross-validation set, and construct the remaining sub-training sets as cross-training sets; The first risk model and the second risk model in the initial group insurance claim settlement model are trained according to the cross-training set, and after the training, cross-validation results of the first risk model and the second risk model on the cross-validation set are obtained.

6. The method for processing a claim risk model according to claim 5, characterized in that: The step of adjusting the model weights of the first risk model and the second risk model according to the obtained cross-validation result to obtain the group insurance claim settlement model includes: Respectively obtaining cross-validation results of the first risk model and the second risk model on each cross-validation set; Calculating the average performance of the first risk model and the second risk model in the cross validation respectively according to the obtained cross validation results; Based on the obtained average performance, model weights of the first risk model and the second risk model are determined, and a group insurance claim model is generated based on the determined model weights.

7. The claim risk model processing method according to claim 1, characterized in that: The step of inputting the target sample into the group insurance claim model to obtain the group insurance claim risk prediction result comprises: Inputting the target sample into the first risk model in the group insurance claim settlement model to obtain a second prediction result; Inputting the target sample into the second risk model in the group insurance claim settlement model to obtain a third prediction result; According to the model weights of the first risk model and the second risk model, the second prediction result and the third prediction result are weighted calculated to obtain a group claim risk prediction result.

8. A claim risk model processing device, characterized in that: include: A training set acquisition module, used to acquire a first training set, wherein the first training set includes original samples and risk labels of multiple group insurance customers; A first training module is used to train a first initial risk model according to the first training set to obtain a first risk model, and obtain the risk prediction result of the first risk model for each group insurance customer; A second training module, used to construct a second training set according to the obtained risk prediction result and the first training set, and train a second initial risk model according to the second training set to obtain a second risk model; An integrated training module, configured to perform integrated training on the first risk model and the second risk model based on an integrated algorithm and the second training set to obtain an initial group insurance claims model; A cross-validation module, used to cross-validate the first risk model and the second risk model in the initial group insurance claim settlement model, and adjust the model weights of the first risk model and the second risk model according to the obtained cross-validation results to obtain a group insurance claim settlement model; The risk prediction module is used to obtain a target sample of target group insurance customers and input the target sample into the group insurance claim model to obtain a group insurance claim risk prediction result.

9. A computer device, characterized in that: It comprises a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the claim risk model processing method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the claim risk model processing method according to any one of claims 1 to 7.

Images