CN117952756A

CN117952756A - Vehicle insurance pricing method, device, equipment and medium

Info

Publication number: CN117952756A
Application number: CN202410059646.0A
Authority: CN
Inventors: 李明; 王珏华
Original assignee: Great Wall Motor Co Ltd
Current assignee: Great Wall Motor Co Ltd
Priority date: 2024-01-15
Filing date: 2024-01-15
Publication date: 2024-04-30

Abstract

The application provides a method, a device, equipment and a medium for pricing vehicle insurance, wherein the method is applied to the field of vehicle insurance, and comprises the following steps: acquiring risk characteristic data in a first time period of an insurance application vehicle; inputting the risk characteristic data into a first prediction model, and outputting a risk assessment result; the risk assessment results include risk transition probabilities that characterize a probability of the application vehicle transitioning between different risk levels; inputting the risk assessment result and the risk characteristic data into a second prediction model, and outputting predicted odds and predicted odds intensity; and determining the premium of the insurance application vehicle according to the predicted pay frequency and the predicted pay intensity. The method can provide more accurate and personalized insurance premium pricing.

Description

Vehicle insurance pricing method, device, equipment and medium

Technical Field

The present application relates to the field of automotive insurance, and more particularly, to a method, apparatus, device and medium for pricing automotive insurance in the field of automotive insurance.

Background

Vehicle insurance is an important financial service aimed at providing reimbursement to the vehicle owner to address potential losses due to unforeseen events such as accidents. At present, in the field of automobile insurance, traditional automobile insurance pricing methods mainly depend on statistical methods and basic risk assessment models, and the methods have limitations, so that automobile insurance cost pricing is not accurate enough and actual risk demands of individual automobile owners cannot be met.

Disclosure of Invention

The application provides a vehicle insurance pricing method, device, equipment and medium, which can provide more accurate and personalized vehicle insurance premium pricing.

In a first aspect, a method of vehicle insurance pricing is provided, the method comprising:

Acquiring risk characteristic data in a first time period of an insurance application vehicle;

Inputting the risk characteristic data into a first prediction model, and outputting a risk assessment result; the risk assessment result comprises risk transition probabilities, wherein the risk transition probabilities represent the probability of transition of the insurance application vehicle between different risk levels;

inputting the risk assessment result and the risk characteristic data into a second prediction model, and outputting predicted odds and predicted odds intensity;

and determining the premium of the insurance application vehicle according to the predicted pay frequency and the predicted pay intensity.

In the technical scheme, more accurate and personalized car insurance premium pricing is realized by respectively inputting the risk characteristic data into the two-stage prediction models. The first predictive model is intended to evaluate risk transition probabilities, knowing the likelihood that an application vehicle transitions between different risk levels. These risk assessment results are then input along with risk profile data to a second predictive model that can be used to accurately predict the odds and intensity. By combining the two-stage model, not only the accuracy of prediction is improved, but also the dynamic change of various risk factors is considered, so that the premium pricing is more in accordance with the unique risk profile of each vehicle, and personalized pricing is realized.

With reference to the first aspect, in some possible implementations, the first prediction model is a Cox proportional-risk regression model, including a plurality of Cox regression sub-models, where risk level transition relationships corresponding to the risk transition probabilities predicted by the plurality of Cox regression sub-models are different.

With reference to the first aspect and the foregoing implementation manner, in some possible implementation manners, inputting the risk feature data into a first prediction model, outputting a risk assessment result includes:

Inputting the risk characteristic data into a Cox proportion risk regression model, and determining the current risk level according to the risk characteristic data;

Determining at least one target Cox regression sub-model in the plurality of Cox regression sub-models according to the current risk level;

processing the risk characteristic data through the at least one target Cox regression sub-model, and outputting a risk assessment result; the risk assessment result comprises risk transition probabilities predicted by the at least one target Cox regression sub-model respectively.

In the technical scheme, the Cox proportional risk regression model is adopted as the first prediction model, is particularly good at processing time dependency data, can capture the change trend of risks along with time, and has high flexibility and practicability. The Cox proportional-risk regression model is constructed as a multi-state Cox model, and comprises a plurality of Cox regression sub-models, when predicting and analyzing the transition probabilities of the insuring vehicles between different risk levels, various possible risk transition paths can be more accurately identified and predicted, and more comprehensive and dynamic risk assessment is provided, so that more reasonable premium strategies can be formulated more effectively.

With reference to the first aspect and the foregoing implementation manner, in some possible implementation manners, the second prediction is a generalized linear model, including a pay frequency sub-model and a pay strength sub-model, where the pay frequency sub-model is used to predict a pay frequency of the insurance vehicle in a second period of time, and the pay strength sub-model is used to predict a pay strength of the insurance vehicle in the second period of time.

With reference to the first aspect and the foregoing implementation manner, in some possible implementation manners, the risk level includes a risk frequency level of a pay frequency dimension and a risk intensity level of a pay intensity dimension, the first prediction is a Cox proportional risk regression model, including a pay frequency Cox regression sub-model and a pay intensity Cox regression sub-model, and the risk evaluation result includes a risk frequency conversion probability and a risk intensity conversion probability;

inputting the risk characteristic data into a first prediction model, and outputting a risk assessment result, wherein the method comprises the following steps:

Inputting the risk characteristic data into the pay frequency Cox regression sub-model, and outputting the risk frequency conversion probability of conversion between different risk frequency grades;

And inputting the risk characteristic data into the odds intensity Cox regression sub-model, and outputting the risk intensity conversion probability of conversion between different risk intensity levels.

With reference to the first aspect and the foregoing implementation manner, in some possible implementation manners, inputting the risk assessment result and the risk feature data into a second prediction model, and outputting a predicted odds and a predicted odds strength includes:

inputting the risk frequency conversion probability and the risk characteristic data in the risk assessment result into the pay frequency submodel, and outputting predicted pay frequency;

and inputting the risk intensity conversion probability and the risk characteristic data in the risk assessment result into the pay intensity sub-model, and outputting predicted pay intensity.

In the technical scheme, the second prediction model adopts the generalized linear model, the generalized linear model comprises sub-models of the pay frequency and the pay intensity, the generalized linear model is adopted to predict the pay frequency and the total pay amount of the insurance vehicle in a certain future time, the accuracy and the flexibility of pay prediction are further enhanced, and the method can flexibly adapt to different types of risk distribution and provide accurate pay prediction according to dynamic risk assessment and risk characteristic data provided by the Cox proportional risk regression model. By combining the time dependency analysis of the Cox proportional risk regression model and the statistical prediction capability of the generalized linear model, a more scientific, fine and highly personalized strategy can be provided for vehicle risk pricing, and the risk can be managed and the premium strategy can be optimized more effectively.

With reference to the first aspect and the foregoing implementation manner, in some possible implementation manners, the risk characteristic data includes one or more of vehicle information, driver information, geographic information, historical payment information, and historical violation information.

In the technical scheme, a set of comprehensive risk assessment data set is provided, and the risk condition of the vehicle can be deeply analyzed, so that the potential risk of paying claims is better understood and predicted, and the accuracy and reliability of risk assessment are ensured.

In a second aspect, there is provided a vehicle insurance pricing device, the device comprising:

The acquisition module is used for acquiring risk characteristic data in a first time period of the insurance application vehicle;

The first prediction module is used for inputting the risk characteristic data into a first prediction model and outputting a risk assessment result; the risk assessment result comprises risk transition probabilities, wherein the risk transition probabilities represent the probability of transition of the insurance application vehicle between different risk levels;

The second prediction module is used for inputting the risk assessment result and the risk characteristic data into a second prediction model and outputting predicted odds and predicted odds;

And the pricing module is used for determining the premium of the insurance application vehicle according to the predicted pay frequency and the predicted pay intensity.

With reference to the second aspect, in some possible implementations, the first prediction is a Cox proportional-risk regression model, including a plurality of Cox regression sub-models, where risk level transition relationships corresponding to the risk transition probabilities predicted by the plurality of Cox regression sub-models are different.

With reference to the second aspect and the foregoing implementation manner, in some possible implementation manners, the foregoing first prediction module is specifically configured to:

With reference to the second aspect and the foregoing implementation manner, in some possible implementation manners, the second prediction is a generalized linear model, including a pay frequency sub-model and a pay strength sub-model, where the pay frequency sub-model is used to predict a pay frequency of the insurance vehicle in a second period of time, and the pay strength sub-model is used to predict a pay strength of the insurance vehicle in the second period of time.

With reference to the second aspect and the foregoing implementation manner, in some possible implementation manners, the risk level includes a risk frequency level of a pay frequency dimension and a risk intensity level of a pay intensity dimension, the first prediction is a Cox proportional risk regression model, including a pay frequency Cox regression sub-model and a pay intensity Cox regression sub-model, and the risk evaluation result includes a risk frequency conversion probability and a risk intensity conversion probability;

The first prediction module is specifically configured to:

With reference to the second aspect and the foregoing implementation manner, in some possible implementation manners, the foregoing second prediction module is specifically configured to:

With reference to the second aspect and the foregoing implementation manner, in some possible implementation manners, the risk characteristic data includes one or more of vehicle information, driver information, geographic information, historical payment information, and historical violation information.

In a third aspect, an electronic device is provided that includes a memory and a processor. The memory is for storing executable program code and the processor is for calling and running the executable program code from the memory for causing the electronic device to perform the method of the first aspect or any one of the possible implementations of the first aspect.

In a fourth aspect, there is provided a computer program product comprising: computer program code which, when run on a computer, causes the computer to perform the method of the first aspect or any one of the possible implementations of the first aspect.

In a fifth aspect, a computer readable storage medium is provided, the computer readable storage medium storing computer program code which, when run on a computer, causes the computer to perform the method of the first aspect or any one of the possible implementations of the first aspect.

Drawings

FIG. 1 is a flow chart of a method for pricing insurance of a vehicle according to an embodiment of the present application;

FIG. 2 is a flow chart of a method for pricing insurance of a vehicle according to an embodiment of the present application;

FIG. 3 is an exemplary flow chart of a method for vehicle insurance pricing provided by an embodiment of the application;

FIG. 4 is a schematic diagram of a vehicle insurance pricing device according to an embodiment of the present application;

fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

The technical scheme of the application will be clearly and thoroughly described below with reference to the accompanying drawings. Wherein, in the description of the embodiments of the present application, unless otherwise indicated, "/" means or, for example, a/B may represent a or B: the text "and/or" is merely an association relation describing the associated object, and indicates that three relations may exist, for example, a and/or B may indicate: the three cases where a exists alone, a and B exist together, and B exists alone, and furthermore, in the description of the embodiments of the present application, "plural" means two or more than two.

The terms "first," "second," and the like, are used below for descriptive purposes only and are not to be construed as implying or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature.

It should be noted that, the information (including but not limited to user equipment information, user personal information, etc.), data (including but not limited to data for analysis, stored data, presented data, etc.), and signals related to the embodiments of the present application are all authorized by the user or are fully authorized by the parties, and the collection, use, and processing of the related data is required to comply with the relevant laws and regulations and standards of the relevant countries and regions. For example, the vehicle information, the history reimbursement information, the history violation information, and the like, which are referred to in the present application, are acquired with sufficient authorization.

The present application will be described in detail with reference to specific examples.

Embodiments of the present application provide a vehicle insurance pricing method that may be implemented in dependence on a computer program, and that may be run on a network data acquisition device based on von neumann systems. The computer program may be integrated in the application or may run as a stand-alone tool class application. The vehicle insurance pricing device in the embodiment of the application can be a terminal device, including but not limited to: personal computers, tablet computers, handheld devices, vehicle mounted devices, wearable devices, computing devices, or other processing devices connected to a wireless modem, and the like. Terminal devices in different networks may be called different names, for example: a user equipment, an access terminal, a subscriber unit, a subscriber station, a mobile station, a remote terminal, a mobile device, a user terminal, a wireless communication device, a user agent or user equipment, a cellular telephone, a cordless telephone, a terminal device in a 5G network or a future evolution network, etc. The vehicle insurance pricing device can also be a server, can be an independent physical server, or can be a server cluster or a distributed system formed by a plurality of physical servers, or can be a cloud server for providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, a distribution network (Content Delivery Network, CDN), basic cloud computing services such as big data and an artificial intelligent platform. The vehicle insurance pricing device can also be a system of a server and terminal equipment combination. The embodiment of the application is not limited to this, and is specifically determined based on the actual application environment.

Next, please refer to fig. 1, which is a flowchart of a vehicle insurance pricing method according to an embodiment of the present application, wherein an execution subject is a terminal device. As shown in fig. 1, the method comprises the following steps:

S101, acquiring risk characteristic data in a first time period of an insurance application vehicle.

In particular, by collecting and analyzing relevant risk profile data of an insurance vehicle over a particular period of time, the necessary data basis may be provided for subsequent risk assessment and insurance pricing.

The first time period refers to a particular historical time frame for collecting risk profile data. The risk profile data refers to various information that can characterize the risk condition of the insurance vehicle. The risk characteristic data within the first time period should be able to reflect the current and recent risk conditions of the vehicle, and the length of the first time period may be determined according to actual needs and data availability. For example, the first period of time may be the last year before the vehicle is warranted, or may be the entire period of the vehicle owner's driving history.

In some embodiments, the risk characteristics data may include, but is not limited to, vehicle information, driver information, geographic information, historical reimbursement information, and historical violation information. Exemplary:

vehicle information, which may include a model of the vehicle being covered, age of the vehicle, mileage, vehicle usage properties (commercial use or private use), maintenance history, and the like;

Driver information, which may include driving age, sex, driving experience, driving habit, etc. of the driver of the insurance vehicle;

geographic information, which may include registered areas or daily travel areas of the insurance vehicle, etc.;

historical payment information, which may include the number of past claims made to the vehicle, the amount of claims made, the type of claims made, etc.;

the history violation information may include traffic violation information that occurred in the past of the application vehicle, and the like.

By collecting the risk characteristic data, a set of comprehensive risk assessment data set can be obtained, and the risk condition of the vehicle can be deeply analyzed, so that potential risk of reimbursement is better understood and predicted, and the accuracy and reliability of subsequent risk assessment are ensured.

In some embodiments, the risk feature data is data obtained by preprocessing and feature engineering of original data. Because the original data often contains incomplete, inconsistent or information which cannot be directly applied to the model, the data quality can be improved through preprocessing and feature engineering, and the risk feature data after preprocessing and feature engineering is beneficial to improving the accuracy and effectiveness of the model.

The data preprocessing may include the steps of cleaning the data, processing missing and outliers, converting the data type, and the like. For example, performing data cleansing may include removing duplicate values to ensure uniqueness of the data set, and visualization tools such as box graphs may be used to identify and process outliers. For missing values, deletion, padding (using median, average, mode, or K-nearest neighbor algorithms, etc.) or other variable-based prediction methods may be employed for processing. In addition, it is also possible to convert the classified data into a digital format (e.g., using one-hot encoding) and to convert the date data into a format that is easy to handle.

The purpose of feature engineering is to extract useful information from the raw risk feature data to enhance the predictive capabilities of the model. Feature engineering may include variable selection, feature generation, and creation of interactive items. For example, variable selection may be based on vehicle insurance industry knowledge and statistical methods, such as correlation analysis and stepwise regression, to determine the variables most relevant to vehicle insurance benefits and risks. In terms of feature generation, new features, such as driving experience, may be created based on vehicle insurance industry knowledge and vehicle insurance pricing practices. At the same time, consider interactive terms between variables, such as a combination of driving years and driver age, and create polynomial features to capture the nonlinear relationship. In addition, features related to pricing based on vehicle risk (such as vehicle value, accident history of driver) and time features (such as specific information of date and time) can be extracted. Through the steps, the data quality can be improved, and the accuracy and the effectiveness of the model can be obviously enhanced.

S102, inputting the risk characteristic data into the first prediction model, and outputting a risk assessment result.

Specifically, the collected risk characteristic data are input into a first prediction model, a risk assessment result is output, the risk assessment result comprises risk transition probabilities, and the risk transition probabilities represent the probability that the application vehicle is transited between different risk levels. It will be appreciated that the purpose of the first predictive model is to evaluate the probability that an insurance vehicle will transition from one risk level to another within a certain period of time in the future, which may be the intensity of the payoff and/or the frequency of payoff, to characterize the degree of risk and possible payoff of the insurance vehicle.

The first prediction model may be selected from a time sequence analysis model (such as autoregressive integral moving average), a random forest model, a gradient elevator, a logistic regression model, a Cox proportional-risk regression model, and the like.

In some embodiments, the first prediction model is a Cox proportional-risk regression model, which is also called Cox model or Cox proportional-risk model, and is a statistical method for survival analysis. The Cox proportional hazards regression model can be used to study the time of occurrence of an event (e.g., a vehicle claim) and its relationship to a number of variables (e.g., driver age, vehicle type, etc.).

At the heart of the Cox proportional-risk regression model is a proportional-risk hypothesis, the basic form of which can be expressed as:

h(t,X)＝h₀(t)exp(β₁X₁+β₂X₂+…+β_pX_p)

Where h (t, X) is the instantaneous risk rate (or risk function) at time t and given covariates X. h ₀ (t) is a baseline risk function that is a function of time, describing the instantaneous risk rate when all covariates are zero. X ₁,X₂,…,X_p is the covariates (e.g., age of driver, driving experience, vehicle type, etc.). Beta ₁,β₂,…,β_p is a model parameter representing the relative impact of each covariate on risk ratio.

The Cox proportional hazards regression model can effectively process time-related data, can evaluate time-varying hazards, and captures time-dependent hazards features in risk assessment of vehicle insurance claims. One core feature of the Cox proportional-risk regression model is the proportional-risk assumption that allows the baseline risk function to change over time without explicitly specifying its form, without having to preset the risk function form, with good flexibility and applicability to risk claims where the baseline risk is difficult to estimate accurately. In addition, the Cox proportional hazards regression model may consider multiple covariates (e.g., driver age, drive history, vehicle type, etc.) simultaneously, which helps to fully evaluate various factors that affect risk.

In this embodiment, the Cox proportional hazards regression model is used to predict the probability of a vehicle transitioning from a current risk level to another risk level, i.e., the risk transition probability, over a period of time in the future. The risk transition probabilities provide quantitative information about changes in vehicle risk conditions, help predict risk patterns that may occur in the future, and may help insurance companies identify which vehicles may be at higher claim risk, thereby calculating premium more accurately.

The baseline risk function is a non-parametric part of the Cox proportional-risk regression model, which can be estimated from a training dataset related to historical vehicle insurance reimbursements, is a comprehensive analysis of the training dataset, represents a basic risk level of occurrence of an event (risk transfer) when covariates (risk features) are not considered, and can be estimated from the training dataset using maximum likelihood estimation or other statistical methods.

Covariates are variables in the Cox proportional hazards regression model that reflect various factors that may affect the risk level, using risk characteristic data (e.g., driver age, vehicle type, drive history, etc.) as variables that are used to adjust the baseline risk rate to more accurately estimate the risk transfer probability under certain conditions.

Model parameters represent the relative influence of each covariate on risk and can be understood as the weight of risk characteristic data. Model parameters are estimated on the training data set by statistical methods (e.g., maximum likelihood estimation) that determine how and to what extent covariates affect risk.

It should be noted that, the Cox proportional hazards regression models used in the embodiments of the present application are trained models using training data sets related to historical vehicle insurance reimbursements, so that the models can accurately predict the probability of an insurance application vehicle transitioning from one risk level to another risk level in a certain period of time in the future.

In this embodiment, the Cox proportional hazards regression model may be specifically used to analyze time-to-event data (e.g., claim events) to effectively process and interpret time-dependent hazards; the form of the baseline risk is not required to be specified, so that the method has high adaptability under various different conditions; multiple risk factors, such as driver age, vehicle type, etc., can be considered simultaneously and the combined impact of these factors on risk transition probabilities can be evaluated. Therefore, compared with other models, the Cox proportional risk regression model can describe the evolution of risks in detail along with time and provide deep insight into future risk level changes, and the obtained risk assessment results can more accurately reflect the risk changes of the insurance vehicle along with time. This deep analysis of time-dependent risk features not only helps to better understand and predict the occurrence of a payoff event, but also provides a powerful data support in formulating a more accurate premium strategy.

In some embodiments, the risk level includes a risk frequency level of the pay frequency dimension and a risk intensity level of the pay intensity dimension, the first prediction model is a Cox proportional risk regression model including a pay frequency Cox regression sub-model and a pay intensity Cox regression sub-model, and the risk assessment result includes a risk frequency transition probability and a risk intensity transition probability. Inputting the risk characteristic data into the Cox proportional risk regression model, and outputting a risk assessment result, wherein the method comprises the following steps: inputting the risk characteristic data into a pay frequency Cox regression sub-model, and outputting risk frequency conversion probability of conversion between different risk frequency grades; and inputting the risk characteristic data into the odds intensity Cox regression sub-model, and outputting risk intensity conversion probability of conversion between different risk intensity levels.

Specifically, in this embodiment the Cox proportional hazards regression model is used for two different dimensions: pay frequency and pay strength. In the pay frequency dimension, the risk level is divided into risk frequency levels, reflecting how frequently the claimed vehicle is subject to claims in a certain time. By way of example, the risk frequency levels may be divided into a number of levels such as "0 pay", "low frequency", "medium frequency", and "high frequency".

In the pay intensity dimension, the risk level is divided into risk intensity levels, reflecting the total amount of claims that the insurance vehicle has incurred in a certain period of time. For example, the risk intensity level may be divided into a plurality of levels of "0 pay", "low pay", "medium pay", and "high pay".

The Cox proportion risk regression model is divided into two types of submodels based on the pay frequency dimension and the pay intensity dimension, and the pay frequency Cox regression submodel is used for predicting the transition probability between different risk frequency grades, namely the possibility that an insurance application vehicle is transferred from one pay frequency grade to the other grade; the odds intensity Cox regression sub-model is used to predict the probability of transition between different risk intensity levels, i.e., the likelihood that an insurance vehicle will transition from one level to another at the odds amount. The risk functions defined by the odds Cox regression sub-model and the odds intensity Cox regression sub-model are different in related dimensionalities, but belong to the Cox proportion risk regression model. It can be understood that the Cox proportional-risk regression model in this embodiment includes Cox models with multiple dimension types, each model can be responsible for a prediction task with one dimension type, can obtain comprehensive evaluation about risk frequency and risk intensity, ensures diversity and accuracy of prediction results, and provides data support for predicting future claim risk and potential claim cost more accurately.

S103, inputting the risk assessment result and the risk characteristic data into a second prediction model, and outputting predicted odds and predicted odds intensity.

Specifically, the risk assessment result and the risk characteristic data obtained by using the first prediction model (such as a Cox proportional-risk regression model) are used for further prediction of the pay frequency and pay intensity through the second prediction model.

The second prediction model can be selected from a linear regression model, a poisson regression model, a decision tree, a random forest model, a generalized linear model (Generalized Linear Model, GLM) and the like.

In some embodiments, the second predictive model is a generalized linear model, and the GLM model is a flexible statistical model that can be used to describe the relationship between the dependent variable and one or more independent variables. GLM is a generalization of the linear model, allowing dependent variables to follow exponential family distributions (e.g., normal, binomial, poisson, etc.).

The basic form of the GLM model can be expressed as:

g(E(Y))＝β₀+β₁X₁+β₂X₂+…+β_pX_p)

Wherein Y is a dependent variable (such as the frequency of payment or the intensity of payment); x ₁,X₂,…,X_p is the argument (risk feature data and risk assessment results of Cox model); beta ₀,β₁,β₂,…,β_p is a model parameter, including intercept beta ₀ and slope coefficient beta ₁,β₂,…,β_p; intercept is a constant term in the model parameters that represents the expected value of the dependent variable when all independent variables are zero; the slope coefficient is a coefficient corresponding to the independent variable in the model parameter, and represents the variation of the expected value of the dependent variable every unit of variation of the independent variable; g (·) is a join function that joins the expected value E (Y) of the dependent variable and the linear predictor, converting the linear combination of independent variables into a prediction of the dependent variable.

In this embodiment, the independent variables include risk feature data collected, and risk assessment results output by the first prediction model, where the risk assessment results of the first prediction model provide a situation of future risk level changes of the applied vehicle, and the GLM model can effectively process various types of data by selecting appropriate connection functions and probability distributions (such as poisson distribution, gamma distribution, etc.). Therefore, compared with other models, the GLM model can combine the risk assessment result and the risk characteristic data output by the first prediction model to become an ideal bridge for connecting the risk assessment and the actual premium calculation, personalized risk quantification can be carried out on each insurance application vehicle, more comprehensive risk analysis is provided, and more reliable and accurate prediction of the odds and the predicted odds are obtained.

In some embodiments, the generalized linear model includes a pay frequency sub-model for predicting pay frequencies of the application vehicle over the second period of time and a pay strength sub-model for predicting pay strengths of the application vehicle over the second period of time.

Specifically, the GLM model is divided into two sub-models: and (5) a pay frequency sub-model and a pay strength sub-model. The pay frequency sub-model is used to predict pay frequencies of the insurance application vehicle over a future period of time, which may be a count, such as the number of claims in a year. Considering that the odds are count data, the odds submodel may select poisson distribution or negative binomial distribution as the distribution of the response variable, and may map the linear predictor to the expected value of the odds using a logarithmic connection function.

The pay strength sub-model is used to predict the total amount of pays, i.e., the pay strength, of the insurance application vehicle over a period of time in the future. The payoff amount is a positive value, so the payoff intensity sub-model can select a gamma distribution or a normal distribution; depending on the distribution selected, a logarithmic or identity connection function may be used, with the logarithmic connection function being applied to the gamma distribution and the identity connection being applied to the normal distribution.

It should be noted that, the GLM models used in the embodiments of the present application are all trained models using training data sets related to insurance reimbursement of historical vehicles, so that the models can accurately predict reimbursement frequency and reimbursement strength of the insurance application vehicles in a certain period of time in the future.

In some embodiments, inputting the risk assessment result and the risk feature data into the second prediction model, and outputting the predicted odds and the predicted odds intensity includes:

Inputting the risk frequency conversion probability and the risk characteristic data in the risk assessment result into a pay frequency sub-model, and outputting predicted pay frequency;

and inputting the risk intensity conversion probability and the risk characteristic data in the risk assessment result into the pay intensity sub-model, and outputting the predicted pay intensity.

Specifically, the Cox proportional-risk regression model is divided into two kinds of sub-models based on the pay frequency dimension and pay intensity dimension, and the pay frequency Cox regression sub-model is used for predicting the transition probability among different risk frequency grades to obtain the risk frequency transition probability. The risk frequency conversion probability and the risk characteristic data are input into the pay frequency submodel, so that the predicted pay frequency can be obtained more accurately. And predicting the transition probability among different risk intensity levels by using the odds Cox regression sub-model to obtain the risk intensity transition probability. The risk intensity conversion probability and the risk characteristic data are input into the pay intensity sub-model, so that the predicted pay intensity can be obtained more accurately. Therefore, the risk assessment result in the Cox proportion risk regression model can be fully and reasonably utilized, and the accuracy of the pay frequency and pay strength prediction is improved.

S104, determining the premium of the insurance application vehicle according to the predicted pay frequency and the predicted pay intensity.

Specifically, the predicted odds and the predicted odds are the basis for the calculation of the benefit, and the predicted odds are combined to estimate the total odds cost that may be faced. For example, the expected net premium for the insurance vehicle may be determined by multiplying the odds with the odds strength, and then combining with other fee structures and profit goals, the actual premium for each insurance vehicle is ultimately calculated. The specific calculation method of the premium may be specifically set according to the actual implementation, and is not further limited herein.

In the embodiment of the application, a highly accurate and personalized car insurance premium pricing method is realized by combining the application of a Cox proportional risk regression model and a Generalized Linear Model (GLM). Through the Cox model, the risk characteristics changing along with time can be captured, and the probability of transition between different risk levels in the future can be accurately estimated. When the evaluation result of the Cox model is input into the GLM model, the accuracy of the GLM model on the pay frequency and pay intensity prediction is greatly enhanced. The flexibility of the GLM model enables it to efficiently process data of various distribution types and accurately map the relationship between risk features and predictors through appropriate connection functions. The method not only improves the accurate pricing capability of the vehicle insurance products, but also realizes more personalized premium pricing by considering time dynamics and individual differences, and effectively balances risk management and customer satisfaction.

Next, please refer to fig. 2, which is a flowchart of an exemplary vehicle insurance pricing method according to an embodiment of the present application, wherein an execution body is a terminal device. As shown in fig. 2, the method comprises the following steps:

s201, acquiring risk characteristic data in a first time period of an insurance application vehicle.

Specifically, the step S201 corresponds to the step S101, and will not be described herein.

S202, determining the current risk level according to the risk characteristic data.

Specifically, the Cox proportional risk regression model is a multi-state Cox model, and comprises a plurality of Cox regression sub-models, wherein risk grade conversion relations corresponding to risk conversion probabilities predicted by the Cox regression sub-models are different. By way of example, the Cox regression sub-model a may predict the probability that an applied vehicle will transition from one low frequency risk to a medium frequency risk over a period of time in the future; the Cox regression sub-model B can predict the probability that the applied vehicle will transition from a low frequency risk to a high frequency risk in a certain period of time in the future; the Cox regression sub-model C can predict the probability that the applied vehicle will transition from a high risk of reimbursement to a low risk of reimbursement, and so on, over a period of time in the future. Each Cox regression sub-model is responsible for predicting a risk transition probability corresponding to a risk level transition relationship.

Therefore, after inputting the risk feature data into the Cox proportional-risk regression model, the current risk level of the insuring vehicle needs to be determined first, and after a series of Cox regression sub-models needed to be used by the insuring vehicle are determined, the risk feature data of the insuring vehicle is input into the corresponding Cox regression sub-model. The current risk level of the insurance vehicle may be determined from the risk characteristics data of the insurance vehicle.

S203, determining at least one target Cox regression sub-model in the plurality of Cox regression sub-models according to the current risk level.

Specifically, the target Cox regression sub-model is the Cox regression sub-model that matches the current risk level of the insurance vehicle. By way of example, for example, the current risk level of the insurance vehicle is a medium-risk of reimbursement, then the target Cox regression sub-model may include a Cox regression sub-model that predicts a medium-risk of reimbursement to a low-risk of reimbursement, a Cox regression sub-model that predicts a medium-risk of reimbursement to a 0-risk of reimbursement, and a Cox regression sub-model that predicts a medium-risk of reimbursement to a high-risk of reimbursement. I.e., the target Cox regression sub-model is used to predict the probability that the current risk level of the insurance vehicle will transition to other risk levels.

As shown in fig. 3, in some embodiments, the current risk level of the insuring vehicle may include both a risk intensity level and a risk frequency level. For example, the current risk level of the insurance vehicle may be a medium-risk of reimbursement (risk intensity level) and a low-frequency risk of reimbursement (risk frequency level), and each dimension of the current risk level may correspond to one or more target Cox regression sub-models.

S204, processing risk characteristic data through at least one target Cox regression sub-model, and outputting a risk assessment result.

Specifically, the risk characteristic data of the insuring vehicle is input into the target Cox regression sub-model for processing, and the Cox regression sub-model construction and processing process is consistent with the step S102, which is not described herein. The output risk assessment result comprises risk conversion probabilities which are respectively predicted by at least one target Cox regression sub-model, namely, how many target Cox regression sub-models participate in the prediction process, so that the risk assessment result comprises how many risk conversion probabilities, and each risk conversion probability corresponds to one target Cox regression sub-model.

S205, inputting the risk assessment result and the risk characteristic data into a generalized linear model, and outputting predicted odds and predicted odds intensity.

Specifically, the risk conversion probability and risk characteristic data output by each target Cox regression sub-model are input into the GLM model together, and the predicted odds are calculated. The steps of constructing and processing the GLM model are consistent with the step S103, and will not be described in detail herein.

As shown in fig. 3, in some embodiments, the GLM model may be divided into a pay frequency sub-model and a pay intensity sub-model, wherein the risk conversion probability output by the Cox regression sub-model predicting the risk frequency level is input to the pay frequency sub-model, and the risk conversion probability output by the Cox regression sub-model predicting the risk intensity level is input to the pay intensity sub-model, so as to calculate the final predicted pay frequency and predicted pay intensity, respectively.

S206, determining the premium of the insurance application vehicle according to the predicted pay frequency and the predicted pay intensity.

Specifically, the step S206 is identical to the step S104, and will not be described herein.

In the embodiment of the application, the Cox proportional hazards regression model is a multi-state Cox model and comprises a plurality of Cox regression sub-models, which can provide more comprehensive and dynamic risks assessment, particularly can identify and predict various possible risk transition paths more accurately when predicting and analyzing the transition probabilities of the insuring vehicles between different risk levels, so that more reasonable premium strategies can be formulated more effectively. After the method is combined with a Generalized Linear Model (GLM), accuracy of the GLM model in predicting the pay frequency and the pay intensity is remarkably enhanced, and accurate pricing capacity of vehicle insurance products is improved.

Referring next to fig. 4, fig. 4 is a schematic diagram of a vehicle insurance pricing device according to an embodiment of the present application.

As shown in fig. 4, the apparatus 400 includes:

An acquisition module 410, configured to acquire risk feature data in a first period of time of an insurance application vehicle;

The first prediction module 420 is configured to input the risk characteristic data into a first prediction model, and output a risk assessment result; the risk assessment result comprises risk transition probabilities, wherein the risk transition probabilities represent the probability of transition of the insurance application vehicle between different risk levels;

a second prediction module 430, configured to input the risk assessment result and the risk characteristic data into a second prediction model, and output a predicted odds and a predicted odds strength;

And a pricing module 440 for determining the premium of the insurance application vehicle based on the predicted odds and the predicted odds.

In some embodiments, the first prediction model is a Cox proportional-risk regression model, including a plurality of Cox regression sub-models, where risk level transition relationships corresponding to the risk transition probabilities predicted by the Cox regression sub-models are different.

In some embodiments, the first prediction module 420 is specifically configured to:

In some embodiments, the second prediction model is a generalized linear model, and includes a pay frequency sub-model for predicting pay frequencies of the insurance vehicle in a second period of time, and a pay intensity sub-model for predicting pay intensities of the insurance vehicle in the second period of time.

In some embodiments, the risk level includes a risk frequency level of a pay frequency dimension and a risk intensity level of a pay intensity dimension, the first prediction model is a Cox proportional risk regression model, including a pay frequency Cox regression sub-model and a pay intensity Cox regression sub-model, and the risk assessment result includes a risk frequency transition probability and a risk intensity transition probability;

The first prediction module 420 is specifically configured to:

In some embodiments, the second prediction module 430 is specifically configured to:

In some embodiments, the risk profile data includes one or more of vehicle information, driver information, geographic information, historical reimbursement information, and historical violation information.

The above-described division of the modules in the vehicle insurance pricing device is for illustration only, and in other embodiments, the vehicle insurance pricing device may be divided into different modules as needed to perform all or part of the above-described functions of the vehicle insurance pricing device. The implementation of each module in the vehicle insurance pricing device provided in the embodiment of the application can be in the form of a computer program. The computer program may run on a terminal or a server. Program modules of the computer program may be stored in the memory of the terminal or server. Which when executed by a processor, implements all or part of the steps of the vehicle insurance pricing method described in embodiments of the application.

For example, as shown in fig. 5, the vehicle 500 includes: the system comprises a memory 510 and a processor 520, wherein executable program codes 511 are stored in the memory 510, and the processor 520 is used for calling and executing the executable program codes 511 to execute a vehicle insurance pricing method provided by the embodiment of the application.

In addition, the embodiment of the application also protects a device, which can comprise a memory and a processor, wherein executable program codes are stored in the memory, and the processor is used for calling and executing the executable program codes to execute the vehicle insurance pricing method provided by the embodiment of the application.

In this embodiment, the functional modules of the apparatus may be divided according to the above method example, for example, each functional module may be corresponding to one processing module, or two or more functions may be integrated into one processing module, where the integrated modules may be implemented in a hardware form. It should be noted that, in this embodiment, the division of the modules is schematic, only one logic function is divided, and another division manner may be implemented in actual implementation.

In the case of dividing each functional module by corresponding each function, the apparatus may further include an acquisition module, a first prediction module, a second prediction module, a pricing module, and the like. It should be noted that, all relevant contents of each step related to the above method embodiment may be cited to the functional description of the corresponding functional module, which is not described herein.

It should be appreciated that the apparatus provided in this embodiment is used to perform a vehicle insurance pricing method as described above, and thus the same effects as those of the implementation method described above can be achieved.

In case of an integrated unit, the apparatus may comprise a processing module, a memory module. Wherein, when the device is applied to a vehicle, the processing module can be used for controlling and managing the action of the vehicle. The memory module may be used to support the vehicle in executing associated program code, etc.

Wherein the processing module may be a processor or controller that may implement or perform the various exemplary logic blocks, modules and circuits described in connection with this disclosure. A processor may also be a combination of computing functions, including for example one or more microprocessors, digital Signal Processing (DSP) and microprocessor combinations, etc., and a memory module may be a memory.

In addition, the device provided by the embodiment of the application can be a chip, a component or a module, wherein the chip can comprise a processor and a memory which are connected; the memory is used for storing instructions, and when the processor calls and executes the instructions, the chip can be caused to execute the vehicle insurance pricing method provided by the embodiment.

The present embodiment also provides a computer-readable storage medium having stored therein computer program code which, when run on a computer, causes the computer to perform the above-described related method steps to implement a vehicle insurance pricing method provided by the above-described embodiments.

The present embodiment also provides a computer program product which, when run on a computer, causes the computer to perform the above-described related steps to implement a vehicle insurance pricing method provided by the above-described embodiments.

The apparatus, the computer readable storage medium, the computer program product, or the chip provided in this embodiment are used to execute the corresponding method provided above, and therefore, the advantages achieved by the apparatus, the computer readable storage medium, the computer program product, or the chip can refer to the advantages of the corresponding method provided above, which are not described herein.

It will be appreciated by those skilled in the art that, for convenience and brevity of description, only the above-described division of the functional modules is illustrated, and in practical application, the above-described functional allocation may be performed by different functional modules according to needs, i.e. the internal structure of the apparatus is divided into different functional modules to perform all or part of the functions described above.

In the embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of modules or units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another apparatus, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.

The foregoing is merely illustrative of the present application, and the present application is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present application. Therefore, the protection scope of the application is subject to the protection scope of the claims.

Claims

1. A method of vehicle insurance pricing, the method comprising:

inputting the risk characteristic data into a first prediction model, and outputting a risk assessment result; the risk assessment results include risk transition probabilities that characterize a probability of the application vehicle transitioning between different risk levels;

2. The method of claim 1, wherein the first predictive model is a Cox proportional-risk regression model comprising a plurality of Cox regression sub-models, wherein the risk transition probabilities predicted by each of the plurality of Cox regression sub-models correspond to different risk level transition relationships.

3. The method of claim 2, wherein inputting the risk feature data into a first predictive model, outputting a risk assessment result, comprises:

Processing the risk characteristic data through the at least one target Cox regression sub-model, and outputting a risk assessment result; the risk assessment results include risk transition probabilities predicted by each of the at least one target Cox regression sub-model.

4. The method of claim 1, wherein the second predictive model is a generalized linear model including a pay frequency sub-model for predicting pay frequencies of the application vehicle over a second period of time and a pay strength sub-model for predicting pay strengths of the application vehicle over the second period of time.

5. The method of claim 4, wherein the risk level comprises a risk frequency level of a pay frequency dimension and a risk intensity level of a pay intensity dimension, the first predictive model is a Cox proportional risk regression model comprising a pay frequency Cox regression sub-model and a pay intensity Cox regression sub-model, and the risk assessment results comprise a risk frequency transition probability and a risk intensity transition probability;

Inputting the risk characteristic data into the odds frequency Cox regression sub-model, and outputting the risk frequency conversion probability of conversion between different risk frequency levels;

and inputting the risk characteristic data into the odds intensity Cox regression sub-model, and outputting the risk intensity transition probability of transition between different risk intensity levels.

6. The method of claim 5, wherein inputting the risk assessment results and the risk characteristic data into a second predictive model outputs a predicted odds and predicted odds intensity, comprising:

7. The method of claim 1, wherein the risk characteristics data includes one or more of vehicle information, driver information, geographic information, historical reimbursement information, and historical violation information.

8. A vehicle insurance pricing device, the device comprising:

The first prediction module is used for inputting the risk characteristic data into a first prediction model and outputting a risk assessment result; the risk assessment results include risk transition probabilities that characterize a probability of the application vehicle transitioning between different risk levels;

And the pricing module is used for determining the premium of the insurance application vehicle according to the predicted odds and the predicted odds.

9. An electronic device, the electronic device comprising:

A memory for storing executable program code;

a processor for calling and running the executable program code from the memory, causing the vehicle to perform the method of any one of claims 1 to 7.

10. A computer readable storage medium, characterized in that the computer readable storage medium stores a computer program which, when executed, implements the method according to any of claims 1 to 7.