Risk Model Construction Method, Device, Equipment and Medium Based on Similar Vehicle Models

Abstract

The present application provides a method, device, equipment and storage medium for constructing a risk model based on similar vehicle models, relating to the field of data processing. The method of the present application first establishes an initial prediction model related to the driver factor through the risk data of the driver factor of similar vehicle models, avoiding the problem that a new vehicle model cannot establish a risk prediction model due to lack of risk data, thus realizing the risk prediction of the new vehicle model in advance and improving the timeliness of the risk prediction of the new vehicle model. By using the risk data residuals of the vehicle risk data and the predicted risk data, the vehicle factor and its corresponding risk weight, the model parameters of the initial prediction model are optimized to generate a risk prediction model, shortening the establishment time of the risk prediction model of the target vehicle model and improving the risk prediction accuracy of the risk prediction model at the same time.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular, to a method, apparatus, device, and medium for constructing a risk model based on similar vehicle models. Background Art

[0002] In recent years, the popularization of new energy vehicles has brought many challenges to the auto insurance industry. Compared with the original traditional vehicles, the popularization of new energy vehicles has brought many new vehicle models. At the same time, the driving and control methods of new energy vehicles are also very different from those of traditional vehicles. For example, the driving speed and starting speed are often higher than expected, and the forced energy recovery and single-pedal mode may cause discomfort to drivers.

[0003] The above factors have put forward higher requirements for auto insurance pricing. However, at present, there are still deficiencies in the assessment of the risks of new energy vehicles. The main factor is the lack of data accumulation, resulting in low accuracy of risk prediction for new vehicle models.

[0004] Therefore, how to solve the problem of low accuracy of risk prediction for current new vehicle models has become an urgent technical problem to be solved. Summary of the Invention

[0005] The present application provides a method, apparatus, device, and storage medium for constructing a risk model based on similar vehicle models, aiming to improve the accuracy of risk prediction for new vehicle models.

[0006] In a first aspect, the present application provides a method for constructing a risk model based on similar vehicle models. The method for constructing a risk model based on similar vehicle models includes the following steps:

[0007] Establish an initial prediction model based on the driver factor risk data of similar vehicle models;

[0008] Predict the predicted risk data of the target vehicle model based on the initial prediction model;

[0009] Collect the vehicle risk data of the target vehicle model in each of the preset periods based on at least one preset period, and calculate the risk data residual between the vehicle risk data and the predicted risk data;

[0010] Iteratively optimize the initial prediction model based on the preset vehicle factor, the risk data residual, and the risk adjustment weight corresponding to the vehicle risk data to generate a risk prediction model for the target vehicle model.

[0011] In a second aspect, the present application further provides a device for constructing a risk model based on similar vehicle models. The device for constructing a risk model based on similar vehicle models includes:

[0012] A model establishment module, configured to establish an initial prediction model based on the driver factor risk data of similar vehicle models;

[0013] A risk prediction module, configured to predict predicted risk data of a target vehicle model based on the initial prediction model;

[0014] A residual calculation module, configured to collect vehicle risk data of the target vehicle model within each of the at least one preset period, and calculate a risk data residual between the vehicle risk data and the predicted risk data;

[0015] A model parameter optimization module, configured to iteratively optimize the initial prediction model based on a preset vehicle factor, the risk data residual, and a risk adjustment weight corresponding to the vehicle risk data, to generate a risk prediction model of the target vehicle model.

[0016] In a third aspect, the present application further provides a computer device, including a processor, a memory, and a computer program stored on the memory and executable by the processor. When the computer program is executed by the processor, the steps of the method for constructing a risk model based on similar vehicle models as described above are implemented.

[0017] In a fourth aspect, the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for constructing a risk model based on similar vehicle models as described above are implemented.

[0018] The present application provides a method, apparatus, device, and storage medium for constructing a risk model based on similar vehicle models. The method of the present application includes establishing an initial prediction model based on the risk data of the driver factor of similar vehicle models; predicting predicted risk data of a target vehicle model based on the initial prediction model; collecting vehicle risk data of the target vehicle model within each of the at least one preset period, and calculating a risk data residual between the vehicle risk data and the predicted risk data; iteratively optimizing the initial prediction model based on a preset vehicle factor, the risk data residual, and a risk adjustment weight corresponding to the vehicle risk data, to generate a risk prediction model of the target vehicle model. By the above method, the present application first establishes an initial prediction model related to the driver factor through the risk data of the driver factor of similar vehicle models, avoiding the problem that a new vehicle model cannot establish a risk prediction model due to lack of risk data, thereby realizing the risk prediction of the new vehicle model in advance and improving the timeliness of the risk prediction of the new vehicle model. By using the risk data residual between the vehicle risk data and the predicted risk data, the vehicle factor, and its corresponding risk weight, the model parameters of the initial prediction model are optimized to generate a risk prediction model, shortening the establishment time of the risk prediction model of the target vehicle model and improving the risk prediction accuracy of the risk prediction model. Description of the Drawings

[0019] To more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0020] Figure 1 It is a schematic flowchart of the first embodiment of a method for constructing a risk model based on similar vehicle models provided by the present application;

[0021] Figure 2 It is a schematic flowchart of the second embodiment of a method for constructing a risk model based on similar vehicle models provided by the present application;

[0022] Figure 3 It is a schematic structural diagram of the first embodiment of a device for constructing a risk model based on similar vehicle models provided by the present application;

[0023] Figure 4 It is a schematic block diagram of the structure of a computer device provided by the embodiments of the present application.

[0024] The realization of the purpose of the present application, functional features and advantages will be further described with reference to the embodiments and the accompanying drawings. Specific Embodiments

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts fall within the scope of protection of the present application.

[0026] The flowcharts shown in the accompanying drawings are only illustrative examples, and do not necessarily include all contents and operations / steps, nor do they necessarily need to be executed in the described order. For example, some operations / steps can also be decomposed, combined or partially merged, so the actual execution order may be changed according to the actual situation.

[0027] The following will describe in detail some embodiments of the present application with reference to the accompanying drawings. Without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0028] Please refer to Figure 1 , Figure 1 It is a schematic flowchart of the first embodiment of a method for constructing a risk model based on similar vehicle models provided by the present application.

[0029] As Figure 1As shown, the method for constructing a risk model based on similar vehicle models includes steps S101 to S103.

[0030] S101. Establish an initial prediction model based on the risk data of human-related factors for similar vehicle models;

[0031] In one embodiment, for newly launched vehicle models, the biggest problem in modeling is the lack of historical accident data. In the embodiments of the present application, for human-related risk factors, the historical risk data of historical vehicle models similar to the new target vehicle model is used to screen out the risk data of human-related factors, and an initial prediction model related to human-related factors is established for preliminary risk prediction to fill the blank of the lack of risk prediction data for new vehicle models in the early stage.

[0032] In one embodiment, the embodiments of the present application divide the vehicle accident risk into two parts, one part is the risk related to people, and the other is the risk related to the vehicle. First, for the risk related to people, the risk data of human-related factors of similar vehicle models is used to train and construct an initial prediction model related to human-related factors.

[0033] Specifically, collect the risk data of human-related factors in the historical vehicle risk data of similar vehicle models, including information such as driver age, gender, driving experience, driving habits, etc., and train and establish an initial prediction model based on the risk data of human-related factors to predict the safety risk of the driver of the new vehicle model (target vehicle model).

[0034] Furthermore, based on the historical vehicle risk data of the similar vehicle models, screen out the risk data of the human-related factors, wherein the risk data of the human-related factors includes age, gender, driving experience, driving habits, and health status; based on data preprocessing, perform data processing on the risk data of the human-related factors to obtain model training data; based on the regression analysis algorithm and the model training data, establish the initial prediction model.

[0035] In one embodiment, when using the historical vehicle risk data of similar vehicle models to model human data, first screen the historical vehicle risk data of similar vehicle models, and then process the human-related factors for these historical vehicle insurance risk data, that is, the risk data of human-related factors, including age, gender, driving experience, driving habits, health status, etc.

[0036] In one embodiment, through the data preprocessing process, the collected risk data of human-related factors is sorted, cleaned and processed to ensure the accuracy and integrity of the data. Then, using the regression analysis algorithm, establish a regression prediction model of human-related factors as the initial prediction model to predict the driving risk of the target vehicle model.

[0037] Exemplarily, assuming that a linear regression model is used to predict the accident risk R, and referring to the age A and driving experience E in the risk data of human-related factors, then there is:

[0038] R = β0 + β1·A + β2·E + ε

[0039] Among them, β0, β1, and β2 are model parameters, and ε is the error term.

[0040] In one embodiment, data preprocessing is a crucial step in the data analysis and machine learning processes. It involves cleaning, transforming, and normalizing the raw data to facilitate subsequent modeling and analysis. The common processes of data preprocessing are: removing unique attributes, handling missing values, attribute encoding, data standardization and regularization, feature selection, and principal component analysis. Data preprocessing can ensure the accuracy and integrity of the data, thereby improving the accuracy of model prediction.

[0041] In one embodiment, the regression analysis algorithm is to establish a regression relationship function expression (abbreviated as the regression equation) between the dependent variable and the independent variable based on a large amount of observed data using mathematical statistics methods. Regression analysis is a predictive modeling technique that studies the relationship between the dependent variable (target) and the independent variable (predictor), and this relationship of uncertainty (correlation relationship) between the dependent variable and the independent variable.

[0042] S102. Based on the initial prediction model, predict the prediction risk data of the target vehicle model;

[0043] S103. Based on at least one preset period, collect the vehicle risk data of the target vehicle model within each preset period, and calculate the risk data residual between the vehicle risk data and the prediction risk data;

[0044] In one embodiment, the prediction risk data is the risk prediction value obtained by using the initial prediction model.

[0045] In one embodiment, after adjusting the initial prediction model through the vehicle factor risk data to obtain a risk prediction model for the target vehicle model, more vehicle risk data can be continuously collected, and the latest collected vehicle risk data can be used to adjust the parameters of the risk prediction model (calculate residuals, factor modeling, cleaning, and processing) to continuously optimize and iterate the risk identification effect of the risk prediction model. In this way, as the data is continuously updated, the risk prediction model can also always maintain a high prediction accuracy and effectiveness. In this way, a more accurate and effective risk prediction model can be established.

[0046] In one embodiment, the preset period can be collected according to actual needs. For example, it can be set to six months, that is, the vehicle risk data is collected every six months.

[0047] In one embodiment, the vehicle risk data collected each time can be separate, that is, the data collected within each preset period.

[0048] In another embodiment, the vehicle risk data collected each time can also be combined with the historically collected vehicle risk data to form fusion data, that is, all the vehicle risk data within the range from the start time of the first collection of vehicle risk data to the end time of the current collection of vehicle risk data, and the risk prediction model is iteratively optimized with the fused vehicle risk data.

[0049] In one embodiment, the vehicle risk data includes but is not limited to vehicle age, number of accidents, number of maintenance times, maintenance cost, insurance claim amount, and maintenance items.

[0050] Further, based on the time correspondence, the vehicle risk data at the corresponding time of each of the predicted risk data is determined; the risk data residuals of the predicted risk data and the vehicle risk data at the same corresponding time are calculated.

[0051] In one embodiment, the vehicle risk data refers to the risk data actually observed and collected during the use of the target vehicle model, such as the actual number of accidents, insurance claim amount, etc. The predicted risk data is the risk prediction value obtained by using the initial prediction model.

[0052] In one embodiment, according to the time correspondence, the correspondence between the predicted risk data and the vehicle risk data can be determined. For example, according to the time corresponding to the predicted risk data, the vehicle risk data at the corresponding time can be collected to form a data pair of the predicted risk data and the vehicle risk data, and then the risk data residuals of the predicted risk data and the vehicle risk data at the same corresponding time are calculated.

[0053] Exemplarily, for each observation value, the difference between the predicted value and the actual value is calculated. The risk data residual e i can be calculated by the following formula:

[0054]

[0055] where e i is the risk data residual of the i-th observation value, is the vehicle risk data of the i-th observation value, is the predicted risk data of the i-th observation value.

[0056] In one embodiment, the statistical indicators of the risk data residuals, such as mean, standard deviation, skewness, and kurtosis, etc., can be calculated to evaluate the distribution characteristics of the risk data residuals. Ideally, the residuals are randomly distributed without obvious patterns or trends.

[0057] In one embodiment, if the mean of the risk data residuals is close to zero, it indicates that the initial prediction model has no systematic bias. If the standard deviation of the risk data residuals is small, it indicates that the prediction performance of the initial prediction model is relatively accurate. The outliers or extreme points in the risk data residuals are the points where the model prediction is inaccurate. Therefore, according to the results of the residual distribution, the model parameters can be adjusted to optimize the performance of the initial prediction model and improve the model prediction accuracy.

[0058] S104. Based on the preset following-vehicle factors, the risk data residuals, and the risk adjustment weights corresponding to the vehicle risk data, iteratively optimize the initial prediction model to generate the risk prediction model for the target vehicle model.

[0059] In one embodiment, the following-vehicle factors include vehicle age, vehicle price, vehicle configuration, etc.

[0060] It can be understood that the following-vehicle factors have a significant impact on vehicle risk assessment, and they are directly related to the safety of the vehicle, maintenance costs, accident probability, and the frequency and severity of insurance claims.

[0061] In one embodiment, as the vehicle is used over time, its mechanical performance may decline, increasing the probability of failures, thus affecting risk assessment. Vehicles with a greater age may have a higher risk of failures and accidents due to wear and aging.

[0062] In one embodiment, the condition of the vehicle, including maintenance and repair history, also affects its risk assessment. Safety configurations of the vehicle, such as airbags, anti-lock braking system (ABS), electronic stability control, etc., can reduce the severity of accidents.

[0063] In one embodiment, assisted driving functions and intelligent technologies, such as automatic emergency braking, blind spot monitoring, may also affect risk assessment. The usage frequency and purpose of commercial vehicles and non-commercial vehicles are different, affecting their risk exposure. New energy vehicles (such as electric vehicles) may have different risk factors, such as battery safety, availability of charging facilities, etc.

[0064] It can be understood that the impact of the following-vehicle factors on risk assessment is a dynamically changing process, and data and models need to be continuously updated to adapt to these changes.

[0065] In this embodiment, to address the problem of the lack of risk data in the initial stage of a new vehicle model's launch, this solution uses the initial prediction model established from the driver factor risk data as the basis of the risk model, and then uses the following-vehicle factors to dynamically adjust the model parameters of the initial prediction model, gradually adjusting it to the risk prediction model applicable to the target vehicle model.

[0066] In one embodiment, the specific implementation is divided into three steps: First, the risk data residual between the predicted risk data predicted by the initial prediction model and the vehicle risk data is used as the modeling target of the adjustment model, that is, the risk data residual is used as the basis for adjusting the model parameters, so as to gradually reduce the risk data residual and improve the model prediction accuracy; Second, process the vehicle factors, such as vehicle age, vehicle price, vehicle configuration, etc.; Third, set the risk weights of each vehicle factor to adjust the model parameters of the initial prediction model to generate the final risk prediction model.

[0067] The embodiment of the present application ensures risk control while making full use of the advantages of the available data and the model to improve the accuracy of the model risk prediction.

[0068] Further, obtain the acquisition time length of the vehicle risk data; based on the acquisition time length of the vehicle risk data and the standard time length, determine the risk weight of the vehicle factor.

[0069] In one embodiment, divide the data acquisition time length by one year as the adjustment weight. The shorter the data acquisition time range, the smaller the adjustment amplitude.

[0070] In one embodiment, record and determine the start time and end time of the acquisition of the vehicle risk data, and calculate the actual duration of the data acquisition, that is, the acquisition time length.

[0071] In one embodiment, set a standard time length as a reference, for example, it can be set to one year. The ratio of the acquisition time length of the vehicle risk data to the standard time length is used as the weight adjustment factor. For example, if the acquisition time length of the vehicle risk data is 6 months, the weight adjustment factor is 0.5.

[0072] Exemplarily, determine the risk weight of the vehicle factor according to the weight adjustment factor. The risk weight can be calculated according to the following formula: risk weight = 1 / (weight adjustment factor + 1). If the weight adjustment factor is 0.5, the risk weight is 0.333 (that is, 1 / (0.5 + 1)). Apply the calculated risk weight to the vehicle factor to adjust its weight in the risk assessment model.

[0073] Exemplarily, assume that the acquisition time length of a certain vehicle risk data is 18 months (1.5 years). The standard time length is set to 1 year, and the weight adjustment factor is 1.5, then the risk weight is:

[0074]

[0075] That is, it means that in this model adjustment process, the weight of the vehicle factor will be adjusted to 0.4 to reflect the influence of the data acquisition time length on the risk assessment.

[0076] Further, based on the residuals between the vehicle risk data corresponding to each of the preset periods and the predicted risk data of the current risk prediction model, the follower vehicle factor, and the risk weights of each of the risk factors corresponding to each of the prediction periods, parameter iterative optimization is performed on the risk prediction model to generate the risk prediction model corresponding to each of the preset periods.

[0077] Exemplarily, the iterative optimization algorithm can be an algorithm such as the gradient descent method.

[0078] In this embodiment, by combining the residuals, the follower vehicle factor, and its risk weight, parameter iterative optimization is performed on the risk prediction model to adjust the model parameters, which can minimize the prediction error. For each preset period, an updated risk prediction model is generated, which can ensure that the risk prediction model can reflect the risk characteristics of the target vehicle type within each preset period. Regularly reviewing and adjusting the risk weights and model parameters can ensure the prediction accuracy and effectiveness of the risk prediction model.

[0079] In this embodiment, compared with the traditional vehicle insurance risk modeling which often needs to wait for a period of time to collect enough data to establish a risk model, the method of the present application can use data to model and adjust the risk model as early as possible over time in the case of a new vehicle type lacking modeling data, and can understand the risk characteristics of the new vehicle type earlier, so as to provide more accurate and effective risk prediction results for it when the new vehicle type is just put on the market.

[0080] This embodiment provides a method for constructing a risk model based on similar vehicle types. The method of the present application first establishes an initial prediction model related to the driver factor through the driver factor risk data of similar vehicle types, avoiding the problem that a new vehicle type cannot establish a risk prediction model due to lack of risk data, thereby realizing the risk prediction of the new vehicle type in advance and improving the timeliness of the risk prediction of the new vehicle type. Through the risk data residuals between the vehicle risk data and the predicted risk data, the follower vehicle factor, and its corresponding risk weight, model parameter optimization is performed on the initial prediction model to generate a risk prediction model, shortening the establishment time of the risk prediction model of the target vehicle type and at the same time improving the risk prediction accuracy of the risk prediction model.

[0081] Please refer to Figure 2 , Figure 2 which is a schematic flowchart of the second embodiment of the method for constructing a risk model based on similar vehicle types provided by the present application.

[0082] In this embodiment, as Figure 2 shown, based on the above Figure 1 shown embodiment, before the step S101, it further includes:

[0083] S201. Based on at least one vehicle model similarity parameter and the similarity weight corresponding to each vehicle model similarity parameter, calculate the vehicle model similarity between the target vehicle model and at least one historical vehicle model by weighted calculation;

[0084] In one embodiment, the vehicle model similarity parameters include vehicle price, vehicle type, energy type, and main configuration, etc.

[0085] In one embodiment, for a new vehicle model (target vehicle model), use vehicle model similarity parameters such as vehicle price, vehicle type, energy type, and main configuration to perform similarity matching among historical vehicle models, and find historical vehicle models with high similarity as similar vehicle models of the target vehicle model.

[0086] Exemplarily, vehicle price can be matched according to price ranges. For example, 50,000 - 80,000 is the first gear, 80,000 - 120,000 is the second gear, 120,000 - 150,000 is the third gear, 150,000 - 200,000 is the fourth gear, and so on, to distinguish different vehicles according to price ranges.

[0087] Exemplarily, vehicle types can include passenger vehicles (such as sedans, sports cars, multi-purpose vehicles (MPVs), SUVs (sport utility vehicles), and crossover vehicles, etc.), commercial vehicles (trucks, buses, etc.), motorcycles, etc.

[0088] Exemplarily, energy types can include gasoline, diesel, electricity, natural gas, and hybrid power, etc.

[0089] Exemplarily, main configurations can include engine performance (such as maximum power and maximum torque, etc.), vehicle size and weight, safety configurations (such as anti-lock braking system, electronic stability program, airbags, etc.), driving assistance systems (such as rearview camera, radar, etc.), comfort configurations, information entertainment systems, tire and suspension systems, etc.

[0090] In one embodiment, due to different brands, there are certain differences in vehicle pricing, vehicle configurations, etc. for different vehicle models. However, in the screening of similar vehicle models, different similarity weights can be assigned according to the influence degree of different vehicle model similarity parameters on vehicle risks. For example, vehicle price and main configuration have a greater impact on insurance risks, and relatively larger similarity weights can be assigned. While there are more reference types for vehicle type and energy type, which are mainly used to narrow the screening range and have a relatively smaller impact on risk assessment, relatively smaller similarity weights can be assigned.

[0091] In another embodiment, the weights can also be evenly distributed according to the number of vehicle model similarity parameters. For example, for the four vehicle model similarity parameters of vehicle price, vehicle type, energy type, and main configuration, a similarity weight of one-fourth can be assigned to each vehicle model similarity parameter score, and then weighted summation calculation is performed to obtain the final vehicle model similarity.

[0092] Exemplarily, let P = {p1, p2,..., p n} be the parameter set, w i be the weight of parameter p i , v target，i be the value of the target vehicle model on parameter p i , and v hist，i be the value of the historical vehicle model on parameter p i . The parameter similarity s i can be calculated in the following way:

[0093] s i = f(v target，i , v hist，i )

[0094] where f is a function that may have different forms according to the parameter type. For example:

[0095] For quantitative parameters:

[0096]

[0097] where ∈ is a small constant to avoid division by zero.

[0098] For quantitative parameters:

[0099]

[0100] Therefore, the calculation formula for the vehicle model similarity S is:

[0101]

[0102] In a specific embodiment, assume two vehicle models A (target vehicle model) and B (historical vehicle model), and consider three parameters: horsepower, length, and whether there is an ABS system. The weights are 0.5, 0.3, and 0.2 respectively. The parameter values of vehicle model A are (150 horsepower, 4.5 meters, with ABS), and the parameter values of vehicle model B are (145 horsepower, 4.6 meters, with ABS).

[0103] The similarity calculation of the vehicle model similarity parameters is as follows:

[0104] Horsepower similarity:

[0105]

[0106] Length similarity:

[0107]

[0108] ABS similarity:

[0109] s3 = 1

[0110] Weighted summation gives the vehicle model similarity between vehicle models A and B:

[0111] S = 0.5·s1 + 0.3·s2 + 0.2·s3 = 0.898

[0112] That is, it indicates that the vehicle model similarity between the target vehicle model A and the historical vehicle model B is 0.898.

[0113] S202. Based on the vehicle model similarity, determine at least one of the historical vehicle models as a similar vehicle model of the target vehicle model.

[0114] Further, based on the vehicle model similarity corresponding to each historical vehicle model, obtain a similarity ranking; based on the similarity ranking and a preset screening rule, determine at least one of the historical vehicle models as a similar vehicle model of the target vehicle model; wherein, the preset screening rule includes screening according to the number of similar vehicle models and / or screening according to a similarity threshold.

[0115] In one embodiment, by the above method, the vehicle model similarity between a target vehicle model and multiple historical vehicle models can be calculated, and then the vehicle model with the highest similarity can be selected as the similar vehicle model of the target vehicle model.

[0116] In another embodiment, a similarity threshold (such as 0.9) can also be set, and all historical vehicle models with a vehicle model similarity greater than or equal to the similarity threshold are used as similar vehicle models of the target vehicle model.

[0117] In another embodiment, the vehicle model similarities can also be ranked, and a specified number of historical vehicle models can be designated as similar vehicle models of the target vehicle model in descending order of vehicle model similarity, such as designating the top three historical vehicle models in the similarity ranking as similar vehicle models of the target vehicle model, etc.

[0118] In another embodiment, a certain number of historical vehicle models can first be screened by a similarity threshold, and then a specified number of historical vehicle models can be screened from the screened historical vehicle models as similar vehicle models of the target vehicle model; or a specified number of historical vehicle models can be further screened from the screened historical vehicle models according to a certain ratio, for example, 20% of the historical vehicle models with a vehicle model similarity reaching 0.8 are screened as similar vehicle models.

[0119] Please refer to Figure 3 , Figure 3FIG. 0 is a schematic structural diagram of a first embodiment of a risk model construction device based on similar vehicle models provided by the present application. The risk model construction device based on similar vehicle models is used to execute the foregoing risk model construction method based on similar vehicle models.

[0120] As Figure 3 shown, the risk model construction device 300 based on similar vehicle models includes: a model establishment module 301, a risk prediction module 302, a residual calculation module 303, and a model parameter optimization module 304.

[0121] The model establishment module 301 is used to establish an initial prediction model based on the risk data of the driver factors of similar vehicle models;

[0122] The risk prediction module 302 is used to predict the predicted risk data of the target vehicle model based on the initial prediction model;

[0123] The residual calculation module 303 is used to collect the vehicle risk data of the target vehicle model within each of the at least one preset period, and calculate the risk data residual between the vehicle risk data and the predicted risk data;

[0124] The model parameter optimization module 304 is used to iteratively optimize the initial prediction model based on the preset vehicle factors, the risk data residual, and the risk adjustment weight corresponding to the vehicle risk data, and generate the risk prediction model of the target vehicle model.

[0125] In an embodiment, the risk model construction device 300 based on similar vehicle models further includes a risk weight determination module, including:

[0126] An acquisition duration obtaining unit, configured to obtain the acquisition time length of the vehicle risk data;

[0127] A risk weight determination unit, configured to determine the risk weight of the vehicle factor based on the acquisition time length of the vehicle risk data and the standard time length.

[0128] In an embodiment, the risk model construction device 300 based on similar vehicle models further includes a model iterative optimization module, including:

[0129] A data acquisition unit, configured to collect at least one set of vehicle risk data of the target vehicle model based on at least one preset period;

[0130] A model iteration and optimization unit, configured to perform parameter iteration and optimization on the risk prediction model based on the residuals between the vehicle risk data corresponding to each of the preset periods and the predicted risk data of the current risk prediction model, the following-vehicle factors, and the risk weights corresponding to each of the risk factors in each of the prediction periods, so as to generate the risk prediction models corresponding to each of the preset periods.

[0131] In one embodiment, the risk model construction device 300 based on similar vehicle models further includes a similar vehicle model determination module, including:

[0132] A similarity calculation unit, configured to calculate the vehicle model similarity between the target vehicle model and at least one historical vehicle model by weighted calculation based on at least one vehicle model similarity parameter and the similarity weight corresponding to each of the vehicle model similarity parameters;

[0133] A similar vehicle model determination unit, configured to determine at least one of the historical vehicle models as the similar vehicle model of the target vehicle model based on the vehicle model similarity.

[0134] In one embodiment, the similar vehicle model determination unit includes:

[0135] A similarity ranking sub-unit, configured to obtain a similarity ranking based on the vehicle model similarity corresponding to each of the historical vehicle models;

[0136] A similar vehicle model screening sub-unit, configured to determine at least one of the historical vehicle models as the similar vehicle model of the target vehicle model based on the similarity ranking and a preset screening rule;

[0137] Wherein, the preset screening rule includes screening according to the number of similar vehicle models and / or screening according to a similarity threshold.

[0138] In one embodiment, the model establishment module 301 includes:

[0139] A data screening unit, configured to screen out the driver factor risk data based on the historical vehicle risk data of the similar vehicle models, wherein the driver factor risk data includes age, gender, driving experience, driving habits, and health status;

[0140] A data preprocessing unit, configured to perform data processing on the driver factor risk data based on data preprocessing to obtain model training data;

[0141] A model establishment unit, configured to establish the initial prediction model based on a regression analysis algorithm and the model training data.

[0142] In one embodiment, the residual calculation module 303 includes:

[0143] A data corresponding unit, configured to determine the vehicle risk data at the moment corresponding to each of the predicted risk data based on a time correspondence relationship;

[0144] A residual calculation unit, configured to calculate a risk data residual between the predicted risk data and the vehicle risk data corresponding to the same moment.

[0145] It should be noted that those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described device and each module can refer to the corresponding processes in the foregoing embodiments of the risk model construction method based on similar vehicle models, and will not be elaborated herein.

[0146] The device provided in the above embodiment can be implemented in the form of a computer program, and this computer program can run on a computer device as shown in Figure 4 the following.

[0147] Please refer to Figure 4 , Figure 4 which is a schematic block diagram of the structure of a computer device provided in an embodiment of the present application. This computer device can be a server.

[0148] Referring to Figure 4 , this computer device includes a processor, a memory, and a network interface connected through a system bus. Among them, the memory can include a non-volatile storage medium and an internal memory.

[0149] The non-volatile storage medium can store an operating system and a computer program. This computer program includes program instructions, and when the program instructions are executed, the processor can execute any one of the risk model construction methods based on similar vehicle models.

[0150] The processor is configured to provide computing and control capabilities to support the operation of the entire computer device.

[0151] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When this computer program is executed by the processor, the processor can execute any one of the risk model construction methods based on similar vehicle models.

[0152] This network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art can understand that Figure 4 the structure shown in

[0153] It should be understood that the processor can be a Central Processing Unit (CPU), and the processor can also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0154] Among them, in one embodiment, the processor is used to run a computer program stored in a memory to implement the following steps:

[0155] Based on the risk data of the human factor of similar vehicle models, establish an initial prediction model;

[0156] Based on the initial prediction model, predict the predicted risk data of the target vehicle model;

[0157] Based on at least one preset period, collect the vehicle risk data of the target vehicle model in each preset period, and calculate the risk data residual between the vehicle risk data and the predicted risk data;

[0158] Based on the preset vehicle factor, the risk data residual, and the risk adjustment weight corresponding to the vehicle risk data, perform iterative optimization on the initial prediction model to generate the risk prediction model of the target vehicle model.

[0159] In one embodiment, before the processor implements the iterative optimization of the initial prediction model based on the preset vehicle factor, the risk data residual, and the risk adjustment weight corresponding to the vehicle risk data to generate the risk prediction model of the target vehicle model, it is also used to implement:

[0160] Obtain the acquisition time length of the vehicle risk data;

[0161] Based on the acquisition time length of the vehicle risk data and the standard time length, determine the risk weight of the vehicle factor.

[0162] In one embodiment, after the processor implements the iterative optimization of the initial prediction model based on the preset vehicle factor, the risk data residual, and the risk adjustment weight corresponding to the vehicle risk data to generate the risk prediction model of the target vehicle model, it is also used to implement:

[0163] Collect at least one set of vehicle risk data of the target vehicle model based on at least one preset period;

[0164] Based on the residuals between the vehicle risk data corresponding to each preset period and the predicted risk data of the current risk prediction model, the following vehicle factor, and the risk weights corresponding to each risk factor in each prediction period, perform parameter iterative optimization on the risk prediction model to generate the risk prediction model corresponding to each preset period.

[0165] In one embodiment, before the processor implements establishing an initial prediction model based on the driver factor risk data of similar vehicle models, it is further configured to implement:

[0166] Based on at least one vehicle model similarity parameter and the similarity weight corresponding to each vehicle model similarity parameter, calculate the vehicle model similarity between the target vehicle model and at least one historical vehicle model by weighted calculation;

[0167] Based on the vehicle model similarity, determine at least one of the historical vehicle models as a similar vehicle model of the target vehicle model.

[0168] In one embodiment, when the processor implements determining at least one of the historical vehicle models as a similar vehicle model of the target vehicle model based on the vehicle model similarity, it is configured to implement:

[0169] Obtain a similarity ranking based on the vehicle model similarity corresponding to each historical vehicle model;

[0170] Based on the similarity ranking and a preset screening rule, determine at least one of the historical vehicle models as a similar vehicle model of the target vehicle model;

[0171] Wherein, the preset screening rule includes screening according to the number of similar vehicle models and / or screening according to a similarity threshold.

[0172] In one embodiment, when the processor implements establishing an initial prediction model based on the driver factor risk data of similar vehicle models, it is configured to implement:

[0173] Based on the historical vehicle risk data of the similar vehicle models, screen out the driver factor risk data, wherein the driver factor risk data includes age, gender, driving experience, driving habits, and health status;

[0174] Based on data preprocessing, perform data processing on the driver factor risk data to obtain model training data;

[0175] Based on a regression analysis algorithm and the model training data, establish the initial prediction model.

[0176] In one embodiment, when the processor implements calculating the risk data residual of the vehicle risk data and the predicted risk data, it is used to implement:

[0177] Based on the time correspondence relationship, determine the vehicle risk data at the moments corresponding to each of the predicted risk data;

[0178] Calculate the risk data residual of the predicted risk data and the vehicle risk data corresponding to the same moment.

[0179] In an embodiment of the present application, there is also provided a computer-readable storage medium. The computer-readable storage medium stores a computer program, and the computer program includes program instructions. The processor executes the program instructions to implement any one of the risk model construction methods based on similar vehicle models provided in the embodiments of the present application.

[0180] Among them, the computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiment, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a SmartMedia Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the computer device.

[0181] As mentioned above, the above are only specific implementation manners of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed in the present application can easily think of various equivalent modifications or replacements, and these modifications or replacements should be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A method for constructing a risk model based on similar vehicle models, characterized in that, The method includes: Based on at least one vehicle model similarity parameter and the similarity weight corresponding to each vehicle model similarity parameter, calculating the vehicle model similarity between the target vehicle model and at least one historical vehicle model by weighted calculation; Based on the vehicle model similarity, determining at least one of the historical vehicle models as a similar vehicle model of the target vehicle model; Based on the from-person factor risk data of the similar vehicle models, establishing an initial prediction model; Based on the initial prediction model, predicting the predicted risk data of the target vehicle model; Based on at least one preset period, collecting the vehicle risk data of the target vehicle model in each preset period, and based on the time correspondence, determining the vehicle risk data at the moment corresponding to each predicted risk data, and calculating the risk data residual between the predicted risk data and the vehicle risk data at the same moment; wherein, the vehicle risk data includes the vehicle risk data collected in each preset period or the fusion data formed by combining the vehicle risk data collected each time with the historically collected vehicle risk data; Based on the preset from-vehicle factor, the risk data residual, and the risk adjustment weight corresponding to the vehicle risk data, iteratively optimizing the initial prediction model to generate the risk prediction model of the target vehicle model.

2. The method for constructing a risk model based on similar vehicle models according to claim 1, wherein Before the step of iteratively optimizing the initial prediction model based on the preset from-vehicle factor, the risk data residual, and the risk adjustment weight corresponding to the vehicle risk data to generate the risk prediction model of the target vehicle model, it further includes: Obtaining the collection time length of the vehicle risk data; Based on the collection time length of the vehicle risk data and the standard time length, determining the risk weight of the from-vehicle factor.

3. The method for constructing a risk model based on similar vehicle models according to claim 1, wherein The step of iteratively optimizing the initial prediction model based on the preset from-vehicle factor, the risk data residual, and the risk adjustment weight corresponding to the vehicle risk data to generate the risk prediction model of the target vehicle model includes: Based on the residual between the vehicle risk data corresponding to each preset period and the predicted risk data of the current risk prediction model, the from-vehicle factor, and the risk weight corresponding to each risk factor in each preset period, iteratively optimizing the parameters of the risk prediction model to generate the risk prediction model corresponding to each preset period.

4. The risk model construction method based on similar vehicle models according to claim 1, wherein, The step of determining at least one of the historical vehicle models as a similar vehicle model of the target vehicle model based on the vehicle model similarity includes: Obtaining a similarity ranking based on the vehicle model similarity corresponding to each historical vehicle model; Based on the similarity ranking and a preset screening rule, determining at least one of the historical vehicle models as a similar vehicle model of the target vehicle model; Wherein, the preset screening rule includes screening according to the number of similar vehicle models and / or screening according to a similarity threshold.

5. The method for constructing a risk model based on similar vehicle models according to claim 1, wherein The step of establishing an initial prediction model based on the from-person factor risk data of the similar vehicle models includes: Based on the historical vehicle risk data of the similar vehicle models, screening out the from-person factor risk data, wherein the from-person factor risk data includes age, gender, driving experience, driving habits, and health status; Based on data preprocessing, performing data processing on the from-person factor risk data to obtain model training data; Based on the regression analysis algorithm and the model training data, establish the initial prediction model.

6. A risk model construction device based on similar vehicle models, characterized in that, The risk model construction device based on similar vehicle models includes: A similar vehicle model determination module, configured to calculate the vehicle model similarity between a target vehicle model and at least one historical vehicle model by weighted calculation based on at least one vehicle model similarity parameter and the similarity weight corresponding to each vehicle model similarity parameter; based on the vehicle model similarity, determine at least one of the historical vehicle models as the similar vehicle model of the target vehicle model; A model establishment module, configured to establish an initial prediction model based on the risk data of the human factor of the similar vehicle model; A risk prediction module, configured to predict the predicted risk data of the target vehicle model based on the initial prediction model; A residual calculation module, configured to collect the vehicle risk data of the target vehicle model within each of the at least one preset period, and based on the time correspondence, determine the vehicle risk data at the moment corresponding to each predicted risk data; calculate the risk data residual between the predicted risk data and the vehicle risk data at the same moment; wherein the vehicle risk data includes the vehicle risk data collected within each preset period, or the fusion data formed by combining the vehicle risk data collected each time with the historically collected vehicle risk data; A model parameter optimization module, configured to iteratively optimize the initial prediction model based on a preset vehicle factor, the risk data residual, and the risk adjustment weight corresponding to the vehicle risk data, and generate the risk prediction model of the target vehicle model.

7. A computer device, characterized in that, The computer device includes a processor, a memory, and a computer program stored on the memory and executable by the processor, wherein when the computer program is executed by the processor, the steps of the risk model construction method based on similar vehicle models according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, wherein when the computer program is executed by the processor, the steps of the risk model construction method based on similar vehicle models according to any one of claims 1 to 5 are implemented.

Images