Vehicle insurance claim amount prediction method, system and device under natural disaster condition and medium

Abstract

The invention provides a vehicle insurance claim amount prediction method, system and device under a natural disaster condition and a medium, and the method comprises the steps: obtaining a data set of vehicle insurance claim data; the data set is segmented, the data set is segmented into a training set, a test set and a verification set, and the training set is used as a sample set for model training; a base model and a meta model are constructed, and the base model comprises a time sequence base model and a non-time sequence base model; respectively inputting the training set data into the time sequence base model and the non-time sequence base model for training to obtain a prediction result of the base model; and after the base model training is completed, taking prediction results of the time sequence base model and the non-time sequence base model on the verification set as input, inputting the prediction results into the meta-model, and carrying out prediction model training to obtain a final prediction result. The method can achieve the prediction of the insurance claim amount when a natural disaster occurs in each region.

Description

Technical Field

[0001] The present invention relates to the field of automobile insurance claim amount assessment, and in particular to a method, system, device and medium for predicting automobile insurance claim amount in the event of a natural disaster. Background Art

[0002] With the frequent occurrence of climate change and natural disasters, the amount of auto insurance claims has shown an upward trend year by year, which has brought huge economic pressure to insurance companies on the one hand, and on the other hand has put forward higher requirements for the emergency response capabilities of insurance claims. Therefore, accurately predicting the amount of auto insurance claims has become an important means for insurance companies to reduce operating risks and improve operating efficiency.

[0003] At present, insurance companies generally use manual experience and historical data analysis methods to predict the amount of auto insurance claims, but this method has problems such as incomplete data, insufficient samples, and inaccurate models, resulting in low accuracy and reliability of the prediction results. Therefore, building a plan for predicting the amount of auto insurance claims under natural disasters, using artificial intelligence technology and data analysis methods to improve the accuracy and reliability of predictions has become an important means for insurance companies to deal with the operating risks brought by natural disasters. Considering that the occurrence of natural disasters is related to seasonal time and there are many influencing factors, this method intends to use a mixture of time series and non-time series neural networks to construct an integrated learning prediction model to ensure the reliability of the claim volume prediction. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, an object of the present invention is to provide a method for predicting the amount of auto insurance claims in the event of a natural disaster.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0006] The present invention provides a method for predicting the amount of automobile insurance claims in the event of a natural disaster, comprising:

[0007] Get a dataset of auto insurance claims data;

[0008] Splitting the data set into a training set, a test set and a validation set, wherein the training set is used as a sample set for model training, the ratio of the training set to the data set is 7:10, the ratio of the test set to the data set is 2:10, and the ratio of the validation set to the data set is 1:10;

[0009] Constructing a base model and a meta-model, wherein the base model includes a temporal base model and a non-temporal base model;

[0010] Inputting the training set data into the time series base model and the non-time series base model for training respectively, and obtaining the prediction results of the base models, that is, obtaining intermediate results;

[0011] After the base model training is completed, the prediction results (intermediate results) of the time series base model and the non-time series base model on the verification set are used as input and input into the meta-model to perform prediction model training to obtain the final prediction result.

[0012] In a preferred embodiment, the method for predicting the amount of auto insurance claims under natural disasters, after the step of "obtaining a data set of auto insurance claims data", and before the step of "segmenting the data set into a training set, a test set and a validation set, wherein the training set is used as a sample set for model training, the ratio of the training set to the data set is 7:10, the ratio of the test set to the data set is 2:10, and the ratio of the validation set to the data set is 1:10", further includes:

[0013] The data in the data set is processed, the data is cleaned and sorted, outliers are deleted, and missing values ​​are deleted or filled.

[0014] In a preferred embodiment, the method for predicting the amount of auto insurance claims under natural disaster conditions, after the step of "processing the data in the data set, cleaning and arranging the data, deleting outliers, and deleting or filling missing values", and before the step of "segmenting the data set, segmenting the data set into a training set, a test set, and a validation set, wherein the training set is used as a sample set for model training, the ratio of the training set to the data set is 7:10, the ratio of the test set to the data set is 2:10, and the ratio of the validation set to the data set is 1:10", includes:

[0015] Screening the data in the data set to select multiple features associated with the claim amount from the data in the data set;

[0016] By calculating the Pearson correlation coefficient, the 10 features with the highest correlation with the claim amount were selected as key features;

[0017] The one-hot method is used to encode non-numerical features and convert them into numerical type features.

[0018] In a preferred embodiment, the method for predicting the amount of auto insurance claims under natural disasters, after the step of "using the one-hot method to encode non-numerical features and convert them into numerical type features", and before the step of "segmenting the data set into a training set, a test set and a validation set, wherein the training set is used as a sample set for model training, the ratio of the training set to the data set is 7:10, the ratio of the test set to the data set is 2:10, and the ratio of the validation set to the data set is 1:10", further includes:

[0019] The features are normalized by min-max normalization.

[0020] In a preferred embodiment, the method for predicting the amount of auto insurance claims in the event of a natural disaster, the "building a base model and a meta-model, wherein the base model includes a time series base model and a non-time series base model" specifically includes:

[0021] Construct a base model and a meta-model, wherein the base model includes a time series base model and a non-time series base model, the time series base model includes an RNN time series base model and an LSTM time series base model, the non-time series base model includes an SVM non-time series base model and an SVR non-time series base model, and the meta-model adopts an XGBoost network.

[0022] In a preferred embodiment, the method for predicting the amount of auto insurance claims in the event of natural disasters, after the step of "inputting the training set data into the time series base model and the non-time series base model for training respectively to obtain the prediction results (intermediate results) of the base models", and before the step of "after the base model training is completed, using the prediction results (intermediate results) of the time series base model and the non-time series base model on the validation set as input, inputting them into the meta-model, performing prediction model training, and obtaining the final prediction results", includes:

[0023] After the base model training is completed, the data of the test set is input into the trained RNN time series base model to obtain a first result, and the first result is compared with the actual amount of auto insurance claims corresponding to the training set to obtain a first expected error amount, that is, the data of the test set is input into the trained RNN time series base model to obtain a prediction result of the RNN time series base model, and the prediction result is compared with the actual amount of auto insurance claims corresponding to the training set to obtain the expected error amount of the RNN time series base model;

[0024] Input the data of the test set into the trained LSTM time series base model to obtain a second result, compare the second result with the actual amount of auto insurance claims corresponding to the training set, and obtain a second expected error amount, that is, input the data of the test set into the trained LSTM time series base model to obtain a prediction result of the LSTM time series base model, compare the prediction result with the actual amount of auto insurance claims corresponding to the training set, and obtain the expected error amount of the LSTM time series base model;

[0025] Input the data of the test set into the trained SVM non-time series base model to obtain a third result, compare the third result with the actual amount of auto insurance claims corresponding to the training set, and obtain a third expected error amount, that is, input the data of the test set into the trained SVM non-time series base model to obtain a prediction result of the SVM non-time series base model, compare the prediction result with the actual amount of auto insurance claims corresponding to the training set, and obtain the expected error amount of the SVM non-time series base model;

[0026] Input the data of the test set into the trained SVR non-time series base model to obtain a fourth result, compare the fourth result with the actual amount of auto insurance claims corresponding to the training set, and obtain a fourth expected error amount, that is, input the data of the test set into the trained SVR non-time series base model to obtain a prediction result of the SVR non-time series base model, compare the prediction result with the actual amount of auto insurance claims corresponding to the training set, and obtain an expected error amount of the SVR non-time series base model;

[0027] Evaluate and compare the first expected error amount, the second expected error amount, the third expected error amount and the fourth expected error amount. The smaller the expected error amount, the higher the accuracy, and the prediction result of the optimal base model is obtained. That is to say, evaluate and compare the expected error amount of the RNN time series base model, the expected error amount of the RNN time series base model, the expected error amount of the SVM non-time series base model and the expected error amount of the SVR non-time series base model. The smaller the expected error amount, the higher the accuracy, and the prediction result of the optimal base model is obtained.

[0028] In a preferred embodiment, in the method for predicting the amount of auto insurance claims in the event of a natural disaster, in the step of "evaluating and comparing the first expected error amount, the second expected error amount, the third expected error amount and the fourth expected error amount, the smaller the expected error amount, the higher the accuracy, and obtaining the prediction result of the optimal base model", the optimal prediction result is judged by calculating the root mean square error RMSE, that is, the difference between the predicted value and the true value, the smaller the difference, the more accurate the model.

[0029] The present invention also provides a device for predicting the amount of automobile insurance claims in the event of a natural disaster, the device comprising:

[0030] The data acquisition module is used to obtain the data set of motor insurance claims data;

[0031] A data segmentation module is used to segment the data set into a training set, a test set and a validation set, wherein the training set is used as a sample set for model training, the ratio of the training set to the data set is 7:10, the ratio of the test set to the data set is 2:10, and the ratio of the validation set to the data set is 1:10;

[0032] A model building module, used for building a base model and a meta-model, wherein the base model includes a time-series base model and a non-time-series base model;

[0033] A base model processing module, used for inputting the training set data into the time series base model and the non-time series base model for training respectively, and obtaining the prediction result (i.e., intermediate result) of the base model;

[0034] The meta-model processing module is used to, after the base model training is completed, take the prediction results (intermediate results) of the time series base model and the non-time series base model on the verification set as input, input the meta-model, perform prediction model training, and obtain the final prediction result.

[0035] The present invention also provides a system for predicting the amount of automobile insurance claims in the event of a natural disaster, the system comprising at least one processor; and

[0036] a memory communicatively connected to the at least one processor; wherein,

[0037] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned method for predicting the amount of auto insurance claims in natural disaster situations.

[0038] The present invention also provides a non-volatile computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are executed by one or more processors, the one or more processors can execute the above-mentioned method for predicting the amount of auto insurance claims in natural disaster situations.

[0039] Compared with the prior art, the present invention provides a method, system, device and medium for predicting the amount of auto insurance claims under natural disasters, wherein the method for predicting the amount of auto insurance claims under natural disasters includes: obtaining a data set of auto insurance claims data; dividing the data set into a training set, a test set and a validation set, wherein the training set is used as a sample set for model training, the ratio of the training set to the data set is 7:10, the ratio of the test set to the data set is 2:10, and the ratio of the validation set to the data set is 1:10; constructing a base model and a meta-model, wherein the base model includes a time series base model and a non-time series base model; inputting the training set data into the time series base model and the non-time series base model respectively for training to obtain the prediction results of the base model; after the base model training is completed, the prediction results of the time series base model and the non-time series base model on the validation set are used as input, input into the meta-model, and perform prediction model training to obtain the final prediction results. The present invention can predict the amount of insurance claims in various regions when natural disasters occur, helping auto insurance companies to better plan the allocation of claims business resources to cope with the short-term impact of natural disasters on claims services, and improve service quality and social influence. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] 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 the structures shown in these drawings without paying creative work.

[0041] Figure 1 A schematic flow chart of a method for predicting the amount of auto insurance claims in the event of a natural disaster provided by the present invention.

[0042] Figure 2 This is a schematic diagram of the functional modules of the device for predicting the amount of auto insurance claims in the event of a natural disaster provided by the present invention.

[0043] Figure 3 This is a schematic diagram of the hardware structure of the system for predicting the amount of auto insurance claims in the event of a natural disaster provided by the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0046] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present invention, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or suggesting their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0047] like Figure 1 As shown, the present invention provides a method, system, device and medium for predicting the amount of auto insurance claims in the event of natural disasters. The occurrence of natural disasters is usually related to seasons and geographical regions. In order to accurately predict the amount of auto insurance claims in the event of disasters, it is necessary to consider seasonal effects and many factors such as geographical features. When dealing with such complex features, the accuracy of traditional single neural networks is difficult to guarantee. In order to solve this problem, we construct a hybrid multi-model integrated learning prediction method to achieve accurate prediction of the amount of auto insurance claims in the event of natural disasters.

[0048] The present invention provides a method for predicting the amount of automobile insurance claims in the event of a natural disaster, comprising:

[0049] S100, obtaining a data set of automobile insurance claims data;

[0050] S200, (segmenting the data set) segmenting the data set into a training set, a test set and a validation set, wherein the training set is used as a sample set for model training, the ratio of the training set to the data set is 7:10, the ratio of the test set to the data set is 2:10, and the ratio of the validation set to the data set is 1:10;

[0051] S300, constructing a base model and a meta-model, wherein the base model includes a temporal base model and a non-temporal base model;

[0052] S400, inputting the training set data into the time series base model and the non-time series base model for training respectively, and obtaining the prediction result (intermediate result) of the base model;

[0053] S500, after the base model training is completed, the prediction results (intermediate results) of the time series base model and the non-time series base model on the verification set are used as input and input into the meta-model to perform prediction model training to obtain the final prediction result.

[0054] In the present invention, the data source of the data set can be obtained from the auto insurance claims data of the past three years from the auto insurance companies. The training set is used as a sample set for model training; the validation set is used as a data set for model parameter tuning and verification after the initial model is obtained; the test set is used as a sample data set for model accuracy evaluation test after the final model is obtained. In step "S500, after the base model training is completed, the prediction results of the time series base model and the non-time series base model on the validation set are used as input, input into the meta-model, and the prediction model is trained to obtain the final prediction result." After the meta-model is trained, when natural disasters occur in various regions, the actual data can be input into the trained base model, and the prediction result is finally obtained by the meta-model.

[0055] Preferably, the method for predicting the amount of auto insurance claims under natural disaster conditions provided by the present invention, after the step of “S100, obtaining a data set of auto insurance claims data” and before the step of “S200, segmenting the data set into a training set, a test set and a validation set, wherein the training set is used as a sample set for model training, the ratio of the training set to the data set is 7:10, the ratio of the test set to the data set is 2:10, and the ratio of the validation set to the data set is 1:10”, includes:

[0056] S110, processing the data in the data set, cleaning and sorting the data, deleting outliers, and deleting or filling missing values.

[0057] Here we are doing data preprocessing, cleaning and organizing the data, including deleting outliers, missing values, etc., and converting the data into a format suitable for model training. For outlier processing, a certain threshold range is usually set. Data that is not in this range or obviously does not conform to common sense is considered an outlier, and the sample data needs to be deleted. For example, the claim amount should be a positive number > 0. When there is sample data with a negative claim amount in the sample set, the sample should be deleted. For missing value processing, the way to deal with missing data is to fill it in. When the total amount of data is small or the amount of missing data is not large, the missing values ​​can be handled by filling in the missing values. In this method, we use the median filling method to calculate the median of the sample feature and fill in the missing parts with the median.

[0058] Preferably, the method for predicting the amount of auto insurance claims under natural disaster conditions provided by the present invention, after the step of "S110, processing the data in the data set, cleaning and arranging the data, deleting outliers, and deleting or filling missing values", and before the step of "S200, segmenting the data set, segmenting the data set into a training set, a test set, and a validation set, wherein the training set is used as a sample set for model training, the ratio of the training set to the data set is 7:10, the ratio of the test set to the data set is 2:10, and the ratio of the validation set to the data set is 1:10", includes:

[0059] S120, screening the data in the data set, and screening out a plurality of features associated with the claim amount from the data in the data set;

[0060] S130. By calculating the Pearson correlation coefficient, the 10 features with the highest correlation with the claim amount are selected as key features;

[0061] S140. Use the one-hot method to encode non-numerical features and convert them into numerical type features.

[0062] From the raw data of the dataset, we can preliminarily screen out features related to the amount of claims, including vehicle information, driver information, accident information, natural disaster type, claim time, etc. The ten features preliminarily screened out based on statistics and historical experience include vehicle model, vehicle age, driver age, driving experience, month of natural disaster, natural disaster type, claim time, number of claims, claim amount, and insured amount. The excluded features include: driver gender, natural disaster level, vehicle brand, driving habits, accident type, etc.

[0063] By calculating the Pearson correlation coefficient, the 10 features with the highest correlation with the claim amount were screened out as key features. Among them, the type of natural disaster and the claim time were used as fixed feature variables and did not participate in the feature screening process.

[0064] In statistics, the Pearson product-moment correlation coefficient (also known as PPMCC or PCCs, often represented by r or Pearson's r in the article) is used to measure the correlation (linear correlation) between two variables X and Y, and its value is between -1 and 1. In the field of natural sciences, this coefficient is widely used to measure the degree of correlation between two variables. It was developed by Karl Pearson from a similar but slightly different idea proposed by Francis Galton in the 1880s. This correlation coefficient is also called "Pearson's correlation coefficient r".

[0065] The Pearson correlation coefficient between two variables is defined as the quotient of the covariance and standard deviation between the two variables and can be calculated as follows:

[0066]

[0067] The above formula defines the overall correlation coefficient, which is often represented by the Greek lowercase letter ρ. The Pearson correlation coefficient ρ{X,Y} is used to measure the degree of linear correlation between two variables X and Y. This coefficient is equal to the covariance cov(X,Y) between X and Y divided by the product of their respective standard deviations σX and σY, that is, ρ{X,Y}=cov(X,Y) / (σXσY). Specifically, ρXY=cov(X,Y) / (σX*σY), where cov(X,Y) represents the covariance of X and Y, and the calculation formula is: cov(X,Y)=E((X-μX)(Y-μY)), where σX and σY represent the standard deviations of X and Y respectively, μX and μY represent the means of X and Y respectively, and E represents the expectation.

[0068] Preferably, the method for predicting the amount of auto insurance claims under natural disaster conditions provided by the present invention, after the step of "S140, encoding non-numerical features using a one-hot method and converting them into numerical type features", and before the step of "S200, segmenting the data set, segmenting the data set into a training set, a test set and a validation set, wherein the training set is used as a sample set for model training, the ratio of the training set to the data set is 7:10, the ratio of the test set to the data set is 2:10, and the ratio of the validation set to the data set is 1:10", further includes:

[0069] S150, normalize the features through min-max standardization.

[0070] Preferably, in the method for predicting the amount of automobile insurance claims in the case of natural disasters provided by the present invention, the “S300, constructing a base model and a meta-model, wherein the base model includes a time series base model and a non-time series base model” specifically includes:

[0071] S300, constructing a base model and a meta-model, wherein the base model includes a time series base model and a non-time series base model, the time series base model includes an RNN time series base model and an LSTM time series base model, the non-time series base model includes an SVM non-time series base model and an SVR non-time series base model, and the meta-model adopts an XGBoost network.

[0072] RNN (Recurrent Neural Network) is a special neural network structure designed to process sequence data, such as text, time series data, etc. It enables the network to process variable-length sequence inputs and capture the temporal dependencies in the sequence by introducing recurrent connections in the hidden layer. The basic structure of RNN includes input layer, hidden layer and output layer, where there are connections between neurons in the hidden layer to form a cycle. This structure enables RNN to remember previous information and use it in the generation process of the current output.

[0073] LSTM (Long Short-Term Memory) is a special recurrent neural network (RNN) designed to solve the long-term dependency problem in traditional RNN. LSTM effectively solves the gradient vanishing and gradient exploding problems in traditional RNN by introducing three main components: forget gate, input gate, and output gate, so that it can learn and remember information of long time intervals.

[0074] Support Vector Machine (SVM) is a widely used supervised learning algorithm, mainly used for classification and regression analysis. The core idea of ​​SVM is to achieve the best classification of data by finding a hyperplane to maximize the interval between different categories of data. This hyperplane is called the optimal hyperplane, which is determined by the support vectors, that is, those data points closest to the decision boundary.

[0075] Support Vector Regression (SVR) is a regression analysis method based on Support Vector Machine (SVM). SVR creates a "gap band" on both sides of the linear function with a spacing of ε (tolerance deviation). No loss is calculated for all samples falling into the gap band, and only the support vector will affect the function model. SVR obtains the optimized model by minimizing the total loss and maximizing the gap.

[0076] The features that affect the amount of auto insurance claims are mainly divided into two categories: time series and non-time series. In order to fully consider the influence of the two types of variables, in this method, when constructing the base model, we choose RNN and LSTM, which can better handle time correlation problems, as the time series base model, and choose SVM and SVR, which can solve non-linear correlation problems, as the non-time series base model. The base model constructed in this way has better generalization ability in dealing with time series and non-linear problems. For the meta-model, we choose the XGBoost network with gradient boosting ability. Time series features, such as claim time, month of natural disaster occurrence, non-time series features, such as vehicle model, vehicle age, driver age, driving experience, accident type, natural disaster type, number of claims, claim amount, insured amount, etc.

[0077] Before model training, the model can be initialized, such as setting the hyperparameters required for base model and metamodel training, such as learning rate, regularization parameter, kernel function, etc. The present invention will provide some examples below:

[0078] This example provides examples of the main initialization parameters of the RNN and LSTM models:

[0079] Input_size: The dimension of the input feature is 10

[0080] hidden_size: The dimension of the hidden layer features is initially set to 16

[0081] num_layers: The number of LSTM layers is set to 1.

[0082] bias: Whether to use bias, the default is True.

[0083] batch_first: If True, the first dimension of the input and output will be the batch size. Default is False.

[0084] dropout: The dropout probability between layers except the last layer, the default is 0.

[0085] This embodiment will provide an example of SVM and SVR initialization model parameters:

[0086] kernel: The type of function used in the algorithm, such as RBF (radial basis function).

[0087] gamma: kernel function coefficient, used for RBF, poly and sigmoid kernel functions.

[0088] coef0: kernel function constant value, usually 0.0.

[0089] shrinking: Whether to perform heuristics to reduce the size of the problem and help to solve it faster.

[0090] tol: residual convergence condition, that is, training is stopped when the error term reaches the specified value, set to 0.001.

[0091] cache_size: The buffer size is 200, which is used to limit the amount of calculation.

[0092] verbose: Whether to enable verbose output.

[0093] max_iter: The maximum number of iterations defaults to -1, indicating unlimited.

[0094] This example provides an example of initializing the XGboost model parameters:

[0095] Booster: The parameter is the iterative model, select gbtree (tree-based model)

[0096] silent: The parameter determines whether to output information, the default is 0.

[0097] num_feature: The parameter is the feature dimension 10, which does not need to be set manually, and the model will set it automatically.

[0098] max_depth: (default 6), represents the maximum depth of the tree.

[0099] Preferably, the method for predicting the amount of automobile insurance claims under natural disaster conditions provided by the present invention, after the step of "S400, inputting the training set data into the time series base model and the non-time series base model respectively for training to obtain the prediction results of the base models", and before the step of "S500, after the base model training is completed, using the prediction results of the time series base model and the non-time series base model on the validation set as input, inputting them into the meta-model, performing prediction model training, and obtaining the final prediction results", includes:

[0100] S410, after the base model training is completed, the data of the test set is input into the trained RNN time series base model to obtain a first result, and the first result is compared with the actual motor insurance claim amount corresponding to the training set to obtain a first expected error amount;

[0101] S420, inputting the data of the test set into the trained LSTM time series base model to obtain a second result, and comparing the second result with the actual motor insurance claim amount corresponding to the training set to obtain a second expected error amount;

[0102] S430, inputting the data of the test set into the trained SVM non-time series base model to obtain a third result, and comparing the third result with the actual motor insurance claim amount corresponding to the training set to obtain a third expected error amount;

[0103] S440, inputting the data of the test set into the trained SVR non-time series base model to obtain a fourth result, and comparing the fourth result with the actual automobile insurance claim amount corresponding to the training set to obtain a fourth expected error amount;

[0104] S450, evaluating and comparing the first expected error amount, the second expected error amount, the third expected error amount and the fourth expected error amount. The smaller the expected error amount, the higher the accuracy, and the optimal prediction result of the base model is obtained.

[0105] Here, after the training of the base model is completed, the data of the test set is input into the trained RNN time series base model to obtain the prediction result of the RNN time series base model, and the prediction result is compared with the actual auto insurance claim amount corresponding to the training set to obtain the expected error amount of the RNN time series base model; the data of the test set is input into the trained LSTM time series base model to obtain the prediction result of the LSTM time series base model, and the prediction result is compared with the actual auto insurance claim amount corresponding to the training set to obtain the expected error amount of the LSTM time series base model; the data of the test set is input into the trained SVM non-time series base model to obtain the prediction result of the SVM non-time series base model , compare the prediction result with the actual auto insurance claim amount corresponding to the training set to obtain the expected error amount of the SVM non-time series base model; input the data of the test set into the trained SVR non-time series base model to obtain the prediction result of the SVR non-time series base model, and compare the prediction result with the actual auto insurance claim amount corresponding to the training set to obtain the expected error amount of the SVR non-time series base model; finally, evaluate and compare the expected error amount of the RNN time series base model, the expected error amount of the RNN time series base model, the expected error amount of the SVM non-time series base model and the expected error amount of the SVR non-time series base model. The smaller the expected error amount, the higher the accuracy, and the prediction result of the optimal base model is obtained.

[0106] Preferably, in the method for predicting the amount of auto insurance claims under natural disaster conditions provided by the present invention, in the step of "S450, evaluating and comparing the first expected error amount, the second expected error amount, the third expected error amount and the fourth expected error amount, the smaller the expected error amount, the higher the accuracy, and obtaining the prediction result of the optimal base model", the optimal prediction result is judged by calculating the root mean square error RMSE, that is, the difference between the predicted value and the true value, the smaller the difference, the more accurate the model.

[0107] After the model training is completed, the test set is input into the trained model to obtain the model's prediction results. Then, the prediction results are compared with the true labels of the test set to evaluate whether the model's accuracy meets expectations. Evaluation prediction is a regression problem. The best prediction result can be judged by calculating the root mean square error (RMSE), which is the difference between the predicted value and the true value. The smaller the difference, the more accurate the model.

[0108] The calculation formula of RMSE is:

[0109] Where n is the number of samples, yi is the true value of the i-th sample, is the predicted value of the i-th sample, Σ represents the summation operation, and sqrt represents the square root.

[0110] In this way, a method for predicting the amount of auto insurance claims in disaster situations can be constructed through the above method. The steps are as follows:

[0111] S100, obtaining a data set of automobile insurance claims data;

[0112] S110, processing the data in the data set, cleaning and sorting the data, deleting outliers, and deleting or filling missing values.

[0113] S120, screening the data in the data set, and screening out a plurality of features associated with the claim amount from the data in the data set;

[0114] S130. By calculating the Pearson correlation coefficient, the 10 features with the highest correlation with the claim amount are selected as key features;

[0115] S140. Use the one-hot method to encode non-numerical features and convert them into numerical type features.

[0116] S150, normalize the features through min-max standardization.

[0117] S200, dividing the data set into a training set, a test set and a validation set, wherein the training set is used as a sample set for model training, the ratio of the training set to the data set is 7:10, the ratio of the test set to the data set is 2:10, and the ratio of the validation set to the data set is 1:10;

[0118] S300, constructing a base model and a meta-model, wherein the base model includes a time series base model and a non-time series base model; the “S300, constructing a base model and a meta-model, wherein the base model includes a time series base model and a non-time series base model” specifically includes: S300, constructing a base model and a meta-model, wherein the base model includes a time series base model and a non-time series base model, the time series base model includes an RNN time series base model and an LSTM time series base model, the non-time series base model includes an SVM non-time series base model and an SVR non-time series base model, and the meta-model adopts an XGBoost network.

[0119] S400, inputting the training set data into the time series base model and the non-time series base model for training respectively, and obtaining the prediction result of the base model;

[0120] S410, after the base model training is completed, the data of the test set is input into the trained RNN time series base model to obtain a first result, and the first result is compared with the actual motor insurance claim amount corresponding to the training set to obtain a first expected error amount;

[0121] S420, inputting the data of the test set into the trained LSTM time series base model to obtain a second result, and comparing the second result with the actual motor insurance claim amount corresponding to the training set to obtain a second expected error amount;

[0122] S430, inputting the data of the test set into the trained SVM non-time series base model to obtain a third result, and comparing the third result with the actual motor insurance claim amount corresponding to the training set to obtain a third expected error amount;

[0123] S440, inputting the data of the test set into the trained SVR non-time series base model to obtain a fourth result, and comparing the fourth result with the actual automobile insurance claim amount corresponding to the training set to obtain a fourth expected error amount;

[0124] S450, evaluating and comparing the first expected error amount, the second expected error amount, the third expected error amount, and the fourth expected error amount, the smaller the expected error amount, the higher the accuracy, and obtaining the prediction result of the optimal base model. The optimal prediction result is judged by calculating the root mean square error RMSE, that is, the difference between the predicted value and the true value, the smaller the difference, the more accurate the model.

[0125] S500, after the base model training is completed, the prediction results of the time series base model and the non-time series base model on the verification set are used as input and input into the meta-model to perform prediction model training to obtain the final prediction result.

[0126] The present invention provides a method for predicting the amount of auto insurance claims under natural disasters. After fully considering the time series and nonlinear characteristic factors that affect auto insurance claims, an integrated model with both time series and nonlinear processing capabilities is constructed. This method has two advantages: on the one hand, compared with the traditional single model prediction, the use of integrated learning has better stability and more reliable prediction results; on the other hand, when training the integrated model, the time series nonlinear complex factors of the auto insurance claims scenario are fully considered, and the training is carried out in the form of a combination of time series and nonlinear base models, and then the final prediction is made using XGBoost with gradient boosting performance, thereby improving the universality and prediction accuracy of auto insurance claims prediction. The above method can realize the prediction of the amount of insurance claims in various regions when natural disasters occur, helping auto insurance companies to better plan the allocation of claims business resources to cope with the short-term impact of natural disasters on claims services, and improve service quality and social influence.

[0127] Another embodiment of the present invention further provides a device for predicting the amount of auto insurance claims in the event of a natural disaster. Figure 2 As shown, the device comprises:

[0128] A data acquisition module 11 is used to acquire a data set of motor insurance claims data;

[0129] A data segmentation module 12 is used to segment the data set into a training set, a test set and a validation set, wherein the training set is used as a sample set for model training, the ratio of the training set to the data set is 7:10, the ratio of the test set to the data set is 2:10, and the ratio of the validation set to the data set is 1:10;

[0130] A model building module 13, used to build a base model and a meta-model, wherein the base model includes a time-series base model and a non-time-series base model;

[0131] A base model processing module 14 is used to input the training set data into the time series base model and the non-time series base model for training, so as to obtain the prediction results of the base models;

[0132] The meta-model processing module 15 is used to input the prediction results of the time series base model and the non-time series base model on the verification set as input to the meta-model after the base model training is completed, and to perform prediction model training to obtain the final prediction result.

[0133] The module referred to in the present invention refers to a series of computer program instruction segments that can perform specific functions, which are more suitable for the execution process of the method for predicting the amount of auto insurance claims in natural disaster situations than programs. For the specific implementation methods of each module, please refer to the corresponding method embodiments above, which will not be repeated here.

[0134] Another embodiment of the present invention also provides a system for predicting the amount of auto insurance claims in the event of a natural disaster. Figure 3 As shown, the system 10 includes:

[0135] One or more processors 110 and memory 120, Figure 3 A processor 110 is used as an example for description. The processor 110 and the memory 120 may be connected via a bus or other means. Figure 3 The example of connecting through bus is taken in the following.

[0136] The processor 110 is used to complete various control logics of the system 10, and it can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), a single-chip microcomputer, an ARM (Acorn RISC Machine) or other programmable logic device, a discrete gate or transistor logic, a discrete hardware component, or any combination of these components. In addition, the processor 110 can also be any traditional processor, microprocessor or state machine. The processor 110 can also be implemented as a combination of computing devices, for example, a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors combined with a DSP and / or any other such configuration.

[0137] The memory 120, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer executable programs and modules, such as program instructions corresponding to the method for predicting the amount of auto insurance claims under natural disasters in the embodiment of the present invention. The processor 110 executes various functional applications and data processing of the system 10 by running the non-volatile software programs, instructions and units stored in the memory 120, that is, implementing the method for predicting the amount of auto insurance claims under natural disasters in the above method embodiment.

[0138] The memory 120 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required by at least one function; the data storage area may store data created according to the use of the system 10, etc. In addition, the memory 120 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 120 may optionally include a memory remotely arranged relative to the processor 110, and these remote memories may be connected to the system 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0139] One or more units are stored in the memory 120, and when executed by one or more processors 110, the water supply proportional valve fault detection method in any of the above method embodiments is executed, for example, the above described Figure 1 The method comprises steps S100 to S500.

[0140] An embodiment of the present invention provides a non-volatile computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are executed by one or more processors, for example, to execute the above-described Figure 1 The method comprises steps S101 to S103.

[0141] As an example, non-volatile storage media can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) as external cache memory. By way of illustration and not limitation, RAM can be obtained in many forms such as synchronous RAM (SRAM), dynamic RAM, (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), and direct Rambus RAM (DRRAM). The disclosed memory components or memories of the operating environment described herein are intended to include one or more of these and / or any other suitable types of memory.

[0142] In summary, the present invention provides a method, system, device and medium for predicting the amount of auto insurance claims under natural disasters, wherein the method for predicting the amount of auto insurance claims under natural disasters includes: obtaining a data set of auto insurance claims data; dividing the data set into a training set, a test set and a validation set, wherein the training set is used as a sample set for model training, the ratio of the training set to the data set is 7:10, the ratio of the test set to the data set is 2:10, and the ratio of the validation set to the data set is 1:10; constructing a base model and a meta-model, wherein the base model includes a time series base model and a non-time series base model; inputting the training set data into the time series base model and the non-time series base model for training respectively, and obtaining the prediction result of the base model; after the base model training is completed, the prediction result of the time series base model and the non-time series base model on the validation set is used as input, input into the meta-model, and train the prediction model to obtain the final prediction result. The present invention can realize the prediction of the amount of insurance claims in various regions when natural disasters occur.

[0143] The present invention provides a method for predicting the amount of auto insurance claims under natural disasters. After fully considering the time series and nonlinear characteristic factors that affect auto insurance claims, an integrated model with both time series and nonlinear processing capabilities is constructed. This method has two advantages: on the one hand, compared with the traditional single model prediction, the use of integrated learning has better stability and more reliable prediction results; on the other hand, when training the integrated model, the time series nonlinear complex factors of the auto insurance claims scenario are fully considered, and the training is carried out in the form of a combination of time series and nonlinear base models, and then the final prediction is made using XGBoost with gradient boosting performance, thereby improving the universality and prediction accuracy of auto insurance claims prediction. The above method can realize the prediction of the amount of insurance claims in various regions when natural disasters occur, helping auto insurance companies to better plan the allocation of claims business resources to cope with the short-term impact of natural disasters on claims services, and improve service quality and social influence.

[0144] It should be noted that there is not necessarily a certain order between the above-mentioned steps. A person skilled in the art can understand, based on the description of the embodiments of the present invention, that in different embodiments, the above-mentioned steps may have different execution orders, that is, they may be executed in parallel, may be executed interchangeably, and so on.

[0145] What has been described in this specification and the accompanying drawings includes examples of methods, devices, systems and media that can be provided for predicting the amount of auto insurance claims in natural disaster situations. Of course, it is not possible to describe every conceivable combination of elements and / or methods for the purpose of describing the various features of the present disclosure, but it can be recognized that many other combinations and permutations of the disclosed features are possible. Therefore, it is obvious that various modifications can be made to the present disclosure without departing from the scope or spirit of the present disclosure. In addition, or in the alternative, other embodiments of the present disclosure may be apparent from consideration of the present specification and the accompanying drawings and from the practice of the present disclosure as presented herein. It is intended that the examples set forth in this specification and the accompanying drawings be considered to be illustrative and not restrictive in all respects. Although specific terms are employed herein, they are used in a general and descriptive sense and are not used for limiting purposes.

Claims

1. A method for predicting the amount of auto insurance claims under natural disasters, characterized in that: include: Get a dataset of auto insurance claims data; Splitting the data set into a training set, a test set, and a validation set; Constructing a base model and a meta-model, wherein the base model includes a temporal base model and a non-temporal base model; Inputting the training set data into the time series base model and the non-time series base model respectively for training to obtain intermediate results; The intermediate result is input into the meta-model, and prediction model training is performed to obtain a prediction result.

2. The method for predicting the amount of auto insurance claims under natural disasters according to claim 1 is characterized in that: After the step of obtaining a data set of auto insurance claims data and before the step of dividing the data set into a training set, a test set, and a validation set, the method includes: The data in the said dataset is cleaned and organized, outliers are removed, and missing values ​​are deleted or filled.

3. The method for predicting the amount of auto insurance claims under natural disasters according to claim 2 is characterized in that: After the step of "cleaning and arranging the data in the data set, deleting outliers, and deleting or filling missing values", and before the step of dividing the data set into a training set, a test set, and a validation set, the step further includes: Screening the data in the data set to select multiple features associated with the claim amount from the data in the data set; By calculating the Pearson correlation coefficient, the 10 features with the highest correlation with the claim amount were selected as key features; The one-hot method is used to encode non-numerical features and convert them into numerical type features.

4. The method for predicting the amount of auto insurance claims under natural disasters according to claim 3 is characterized in that: After the step of encoding the non-numerical features using the one-hot method and converting them into numerical type features, the method further includes: The features are normalized by min-max normalization.

5. The method for predicting the amount of auto insurance claims under natural disasters according to claim 4 is characterized by: The base model includes a time series base model and a non-time series base model, the time series base model includes an RNN time series base model and an LSTM time series base model, the non-time series base model includes an SVM non-time series base model and an SVR non-time series base model, and the meta-model adopts an XGBoost network.

6. The method for predicting the amount of auto insurance claims in the event of a natural disaster according to claim 5, characterized in that: After the step of inputting the training set data into the time series base model and the non-time series base model respectively for training to obtain an intermediate result, the method further includes: After the base model training is completed, the data of the test set is input into the trained RNN time series base model to obtain a first result, and the first result is compared with the actual auto insurance claim amount corresponding to the training set to obtain a first expected error amount; Inputting the data of the test set into the trained LSTM time series base model to obtain a second result, and comparing the second result with the actual auto insurance claim amount corresponding to the training set to obtain the second expected error amount; Inputting the data of the test set into the trained SVM non-time series base model to obtain a third result, and comparing the third result with the actual auto insurance claim amount corresponding to the training set to obtain the third expected error amount; Inputting the data of the test set into the trained SVR non-time series base model to obtain a fourth result, and comparing the fourth result with the actual auto insurance claim amount corresponding to the training set to obtain a fourth expected error amount; The first expected error amount, the second expected error amount, the third expected error amount and the fourth expected error amount are evaluated and compared. The smaller the expected error amount, the higher the accuracy, and the optimal prediction result of the base model is obtained.

7. The method for predicting the amount of auto insurance claims in the event of a natural disaster according to claim 6, characterized in that: In the step of evaluating and comparing the first expected error amount, the second expected error amount, the third expected error amount and the fourth expected error amount, the smaller the expected error amount, the higher the accuracy, and obtaining the prediction result of the optimal base model, the optimal prediction result is determined by calculating the root mean square error RMSE to judge the difference between the predicted value and the true value, and the smaller the difference, the more accurate the model.

8. A method for predicting the amount of auto insurance claims under natural disasters, characterized in that: The device comprises: The data acquisition module is used to obtain the data set of motor insurance claims data; A data segmentation module is used to segment the data set into a training set, a test set and a validation set, wherein the training set is used as a sample set for model training, the ratio of the training set to the data set is 7:10, the ratio of the test set to the data set is 2:10, and the ratio of the validation set to the data set is 1:10; A model building module, used for building a base model and a meta-model, wherein the base model includes a time-series base model and a non-time-series base model; A base model processing module, used for inputting the training set data into the time series base model and the non-time series base model for training to obtain an intermediate result; The meta-model processing module is used to input the intermediate results into the meta-model, perform prediction model training, and obtain prediction results.

9. A system for predicting the amount of auto insurance claims in the event of natural disasters, characterized in that: The system includes at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for predicting the amount of motor insurance claims in natural disaster situations as described in any one of claims 1-7.

10. A non-volatile computer-readable storage medium, characterized in that: The non-volatile computer-readable storage medium stores computer-executable instructions, which, when executed by one or more processors, enable the one or more processors to execute the method for predicting the amount of auto insurance claims under natural disaster conditions as described in any one of claims 1-7.

Images