Automatic driving insurance pricing method and system based on machine learning

Abstract

The invention discloses an automatic driving insurance pricing method and system based on machine learning, and the method comprises the steps: collecting data of all intelligent networked automobiles connected to an automatic driving network in a region, carrying out the cleaning, denoising and standardization processing of the data, and constructing a group networked automobile overall dynamic large database; according to geographical location information in the vehicle data, dividing the intelligent networked vehicles into different area groups; analyzing the environment feature information of each regional group, and obtaining a regional risk coefficient; analyzing networking communication data between the vehicles in the regional group, and obtaining an interaction risk coefficient between the vehicles; and obtaining a vehicle individual risk coefficient and a driving behavior adjustment factor according to the vehicle data, and constructing a group networking vehicle insurance pricing model in combination with the regional risk coefficient and the inter-vehicle interaction risk coefficient. According to the insurance pricing method, more accurate insurance pricing is carried out from multiple dimensions by collecting the risk data of the whole group intelligent network connection automobiles in the area and combining the interaction relevance between the automobiles in the area.

Description

Technical Field

[0001] The present invention relates to the field of insurance pricing technology, and in particular to an autonomous driving insurance pricing method and system based on machine learning. Background Art

[0002] With the rapid development of intelligent connected vehicle technology and the gradual popularization of autonomous driving technology, the information exchange between vehicles and between vehicles and infrastructure is becoming more and more frequent, providing a richer and more real-time data source for insurance pricing. Intelligent connected vehicles can collect multi-dimensional information such as vehicle status, driving environment, and driving behavior in real time through on-board sensors and vehicle networking communication technology, providing unprecedented refinement for insurance pricing.

[0003] In the current financial insurance industry, especially in the field of automobile insurance, the traditional pricing model of insurance companies mainly relies on the static risk data of a single vehicle, such as the model, age, driver age and historical accident records of the insured vehicle. Although this pricing method can reflect the basic risk level of the vehicle to a certain extent, it ignores the potential impact of the vehicle's operating environment, dynamic changes in driving behavior and technological progress on risk factors, resulting in limited pricing accuracy and difficulty in fully reflecting the actual risk status of the vehicle.

[0004] Currently, although some insurance companies have begun to try to use data such as vehicle mileage and driving behavior for insurance pricing, these methods often lack comprehensive consideration of the vehicle's operating environment and autonomous driving behavior. For example, there are significant differences in the driving risks of vehicles in congested cities and open countryside, and the responsibility for accidents in autonomous driving mode may be borne by the vehicle manufacturer or software supplier. Therefore, how to combine the vehicle's dynamic operating data, driving environment, autonomous driving status, and historical insurance claims records to build a comprehensive risk assessment model has become an urgent problem to be solved in the insurance industry. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention aims to provide an autonomous driving insurance pricing method based on machine learning to solve the problem that the insurance assessment dimension of the vehicle is single in the prior art, the risk assessment model is not comprehensive enough, resulting in limited pricing accuracy and difficulty in fully reflecting the actual risk status of the vehicle. The specific technical solution is as follows: A method for pricing autonomous driving insurance based on machine learning, comprising the following steps: Collect data from all intelligent connected vehicles connected to the autonomous driving network in the region, clean, denoise, and standardize the collected data, integrate individual vehicle data with regional traffic data, and build a large database of group connected vehicles; Divide smart connected vehicles into different regional groups based on the geographic location information in vehicle data; Analyze the environmental characteristics of each regional group, use spatial statistics and time series analysis methods to identify regional risk hotspots and change trends, and obtain regional risk coefficients; Analyze the communication data between vehicles in the regional group, build a risk model for vehicle interaction, and obtain the risk coefficient for vehicle interaction; The vehicle individual risk coefficient and driving behavior adjustment factor are obtained according to the smart networked car data, and a group networked car insurance pricing model is constructed by combining the regional risk coefficient and the vehicle interaction risk coefficient.

[0006] As a preferred implementation mode: the insurance premium pricing method of the group networked automobile insurance pricing model is: , where β is the weight coefficient of vehicle-to-vehicle interaction risk in the overall risk assessment, which is determined by the insurance company’s strategy and market data.

[0007] As a preferred implementation mode: the method for obtaining the regional risk coefficient is: Among them, j represents different areas, nj represents the number of groups in area j, the accident severity coefficient is based on the accident information entered into the autonomous driving network after the accident, and the traffic congestion impact factor is determined according to the real-time road condition information at the time of the accident.

[0008] As a preferred implementation mode: after obtaining the regional risk coefficient, a machine learning model is established for the regional risk coefficient to perform dynamic analysis and prediction, and the machine learning model is: , where RegionRiskMap t+1 represents the regional risk distribution map at the next moment, ConvLSTM(.) represents the convolutional long short-term memory network model, SpatialFeaturest and TemporalFeaturest represent spatial features and temporal features respectively.

[0009] As a preferred implementation manner: the method for obtaining the vehicle-to-vehicle interaction risk coefficient is: , where m represents different types of vehicle-to-vehicle interactions, and the interaction type risk coefficient, interaction frequency, and interaction distance influencing factors are determined based on the real-time data of smart connected cars while driving.

[0010] As a preferred implementation mode: after obtaining the inter-vehicle interaction risk coefficient, a machine learning model is established for the inter-vehicle interaction risk coefficient to perform dynamic analysis and prediction, and the machine learning model is: , where InteractionRisk m,uv represents the interaction risk between vehicles u and v under interaction type m, GNN(.) represents the graph neural network model, VehicleFeatures u / v represents the feature vector of vehicle u / v, EdgeFeatures uv represents the features of the interaction edge between vehicles u and v, and GraphStructure represents the structural information of the vehicle interaction network.

[0011] As a preferred implementation, the risk assessment method for constructing a group network automobile insurance pricing model is: , Among them, OverallRiskScore represents the overall risk score of the group of connected cars, wi represents the weight of the i-th basic model in the ensemble learning, Model i (.) represents the i-th basic risk assessment model (such as XGBoost, Random Forest, etc.), RegionRisk, VehicleRisk, and InteractionRisk represent the assessment results of regional risk, individual vehicle risk, and interaction risk between vehicles, respectively.

[0012] As a preferred implementation mode: the group networked automobile insurance pricing model is dynamically adjusted according to the intelligent networked automobile data, a dynamic risk assessment model is generated, and the insurance premium pricing is dynamically adjusted according to the dynamic risk assessment model.

[0013] The present invention also provides an autonomous driving insurance pricing system based on machine learning, which applies the insurance pricing method and includes: Data collection module: used to collect driving data of the entire group of intelligent connected vehicles connected to the autonomous driving network in the area; Group data processing module: used to clean, standardize and feature engineer the collected group data to extract features related to insurance pricing; Machine learning model training module: used to form a prediction model based on the extracted group and individual vehicle features, which can predict the probability of insurance claims for vehicles; Insurance pricing module: used to dynamically adjust the vehicle's insurance premium based on the output of the prediction model; Feedback learning module: used to collect insurance claim results, integrate the insurance claim results with the group data and optimize the machine learning model.

[0014] The present invention also provides a storage medium, which is a computer-readable storage medium. A computer program is stored on the storage medium, and the computer program implements the insurance pricing method when executed by a processor.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The insurance pricing method of the present invention is different from the traditional insurance pricing based on the risk data of a single vehicle. It collects the risk data of the entire group of intelligent connected vehicles connected to the autonomous driving network in the area, and combines the interactive correlation between vehicles in the area to perform more accurate insurance pricing from multiple dimensions. At the same time, unlike the current insurance industry's annual fixed static insurance pricing for auto insurance, the present invention can dynamically adjust the vehicle's insurance premium rate, perform dynamic pricing, and track the vehicle's driving conditions through the dynamic monitoring and networking functions of intelligent connected vehicles, thereby avoiding the problem of large insurance pricing errors caused by information lags. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flowchart of the autonomous driving insurance pricing method based on machine learning disclosed in the present invention; DETAILED DESCRIPTION The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0017] The following describes the embodiments of the present disclosure through specific examples, and those skilled in the art can easily understand other advantages and effects of the present disclosure from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0018] Embodiment 1 A pricing method for autonomous driving insurance based on machine learning, such as Figure 1 As shown, the following steps are included: S1: Collect data from all connected smart connected vehicles in the region that are connected to the autonomous driving network, clean, denoise, standardize and feature engineer the collected data, integrate individual vehicle data with regional traffic data, and build a large database of group connected vehicles; Data cleaning and denoising refers to identifying and processing missing data, outliers, and duplicate records. For example, missing speed data can be filled in through interpolation, or abnormal acceleration readings can be identified and processed through box plots. The data standardization process refers to the use of mathematical formulas to standardize the data. The formula is: in, x Represents the original data, x' represents the standardized data, m represents the mean of the data, s Represents the standard deviation of the data.

[0019] Among them, feature engineering refers to: selecting features related to insurance pricing, such as mileage, accident records, etc., and generating new features, such as driving behavior scores, and using algorithms to quantify the driver's driving behavior.

[0020] The processed data is used to train a machine learning model to predict the insurance risk of a vehicle.

[0021] The data processing module processes the data using the following formula: in, D processed For the processed data, D raw is the original collected data, f is the data processing function, i are data processing parameters.

[0022] The data acquisition module also collects the vehicle's real-time carbon emission data and the operating status data of the autonomous driving system.

[0023] S2: Divide the intelligent connected vehicles into different regional groups based on the geographic location information in the vehicle data; Because the surrounding environment, topography, road conditions and other factors of automobiles are relatively similar within a certain area, the actual driving conditions of vehicles in the area can be evaluated more accurately. Therefore, smart connected vehicles are divided into different regional groups based on geographic location information. Cars in the same group have similar group characteristics, which can be used as one of the reference standards for insurance pricing.

[0024] Analysis of risk characteristics of the region where the group of connected cars is located: According to the geographic location information in the vehicle data, the intelligent connected cars are divided into different regional groups. The risk characteristics such as vehicle driving behavior patterns, accident rates, traffic congestion conditions, weather conditions, etc. of each regional group are analyzed to form a regional risk profile. Spatial statistics and time series analysis methods are used to identify regional risk hotspots and changing trends.

[0025] S3: Analyze the environmental characteristics of each regional group, use spatial statistics and time series analysis methods to identify regional risk hotspots and change trends, and obtain regional risk coefficients; At the same time, the communication data between vehicles in the regional group is analyzed, the risk model of vehicle interaction is constructed, and the risk coefficient of vehicle interaction is obtained; The regional risk coefficient calculation formula is: Among them, j represents different areas, nj represents the number of groups in area j, the accident severity coefficient is based on the accident information entered into the autonomous driving network after the accident, and the traffic congestion impact factor is determined according to the real-time road condition information at the time of the accident.

[0026] Furthermore, a machine learning model is established to establish a dynamic evolution model for regional risks and monitor the risk changes in the region in real time. The specific methods are as follows: Among them, RegionRiskMap t+1 represents the regional risk distribution map at the next moment, ConvLSTM(.) represents the convolutional long short-term memory network model, SpatialFeatures t and TemporalFeatures t represent spatial features and temporal features respectively.

[0027] Among them, the interaction risk assessment between group networked vehicles: analyze the communication data between vehicles, evaluate the interaction risk factors such as the coordinated driving ability, following distance, and lane change cooperation between vehicles. Construct an interaction risk model between vehicles to quantify the degree of mutual influence between vehicles.

[0028] The calculation formula of the risk coefficient of vehicle interaction is as follows: Among them, m represents different types of vehicle-to-vehicle interactions (such as following a vehicle, changing lanes, etc.), and the interaction type risk coefficient and interaction distance influencing factor are determined based on the real-time data of the smart connected car during driving.

[0029] Further: Enter the machine learning model calculation, establish a dynamic evolution model for vehicle interaction risk, and change the vehicle interaction risk in real time. The vehicle interaction risk assessment model is as follows: Among them, InteractionRisk m,uv represents the interaction risk between vehicles u and v under interaction type m, GNN(.) represents the graph neural network model, VehicleFeatures u / v represents the feature vector of vehicle u / v, EdgeFeatures uv represents the features of the interaction edge between vehicles u and v, and GraphStructure represents the structural information of the vehicle interaction network.

[0030] Dynamic assessment of vehicle risks can not only obtain real-time information about the vehicle's driving conditions and make real-time adjustments to vehicle premiums, but can also predict the vehicle's garage in the future to a certain extent, helping to optimize pricing accuracy and auto insurance underwriting strategies.

[0031] S4: Obtain vehicle individual risk coefficients and driving behavior adjustment factors based on smart networked car data, and build a group networked car insurance pricing model by combining the regional risk coefficient and the vehicle-to-vehicle interaction risk coefficient.

[0032] Among them, the insurance premium pricing method of the group networked car insurance pricing model is: , where β is the weight coefficient of vehicle-to-vehicle interaction risk in the overall risk assessment, which is determined by the insurance company’s strategy and market data.

[0033] Risk assessment of the overall large model of group connected cars: The overall risk assessment system of the large model of group networked vehicles is constructed by integrating regional risk characteristics, individual vehicle risk characteristics and risk factors of interaction between vehicles. The overall large model is risk assessed using machine learning algorithms (such as deep learning, graph neural networks, etc.). During the assessment process, the interaction and impact between different risk factors are considered to ensure the comprehensiveness and accuracy of the risk assessment.

[0034] In some preferred embodiments, the group networked automobile insurance pricing model can be dynamically adjusted: Based on the overall large-scale model risk assessment results, a multi-factor insurance pricing model is constructed. The model incorporates multiple dimensions such as regional risk coefficient, vehicle individual risk coefficient, driving behavior adjustment factor, and vehicle interaction risk coefficient. The parameters and weights in the insurance pricing model are dynamically adjusted according to real-time updated vehicle data and regional risk changes.

[0035] Through the above steps and formulas, the present invention realizes insurance pricing evaluation based on the overall large-scale model risk data of the group of networked vehicles, providing insurance companies with a more accurate, dynamic and comprehensive risk assessment method.

[0036] In some preferred embodiments, different from the current annual fixed static insurance pricing for auto insurance in the insurance industry, the insurance pricing method in this embodiment also considers the carbon emission effect of smart networked vehicles through the dynamic monitoring and networking function of smart networked vehicles, and can dynamically adjust the insurance premium rate of the vehicle and perform dynamic pricing. The specific calculation method is as follows: A risk assessment algorithm is used to generate a dynamic risk assessment model. The algorithm form is: Among them, Rrisk is the risk assessment value, Ecarbon is the carbon emission effect, Edriving is the autonomous driving behavior effect, Eenvironment is the environmental effect, and a, β, γ are weight coefficients.

[0037] The machine learning model training module uses the following loss function for model training: Among them, L represents the loss function, N represents the number of samples, yi represents the actual label, and ẏi represents the predicted label.

[0038] The group networked car insurance pricing model is dynamically adjusted according to the intelligent networked car data, generates a dynamic risk assessment model, and dynamically adjusts the insurance premium pricing according to the dynamic risk assessment model. The specific steps include: Risk score calculation: A machine learning model is used to analyze the vehicle's real-time driving data and historical accident data to derive a risk score, which reflects the probability of the vehicle having an accident in the future.

[0039] Rate adjustment strategy: Based on the risk score, the following mathematical formula is applied to calculate the insurance premium: Among them, P represents the insurance premium rate, P0 represents the basic rate, R represents the risk assessment result, C represents the carbon emission score, β1 and β2 are adjustment coefficients, which are obtained based on market research and historical claims data analysis.

[0040] Through the above method, insurance premiums can be updated in real time: the insurance pricing module monitors vehicle data in real time, and when a significant change in the risk score or carbon emission score is detected, the insurance premium is automatically recalculated and adjusted to minimize inaccurate vehicle scores caused by information lags.

[0041] The insurance pricing module adjusts insurance pricing through the following formula: Among them, Pinsurance is the adjusted insurance cost, Pbase is the basic insurance cost, △P is the risk adjustment coefficient, and Rrisk is the risk assessment value.

[0042] The present invention also provides an autonomous driving insurance pricing system based on machine learning, which includes the following modules: 1. Data collection module: used to collect driving data of the entire group of intelligent networked vehicles connected to the autonomous driving network in the area, including: speed, acceleration, braking frequency, mileage, vehicle model, vehicle age, vehicle maintenance record, vehicle ownership, insurance type, claim record, displacement, fuel efficiency, driver age, gender, driver's license acquisition time, driving years, historical traffic violation records, accident records, average driving speed, accident time and location, accident type, accident responsibility judgment, etc. of the group of intelligent networked vehicles in the area within the same time period. The data collection module is realized through a variety of sensors and data receivers integrated in the intelligent networked vehicles, including speed sensors, acceleration sensors, GPS positioning systems, brake system sensors, emission monitoring systems, etc. All collected data are uploaded to the central server through the vehicle's communication system for further processing and analysis.

[0043] 2. Data processing module: Used to clean, standardize and feature engineer the collected group data to extract features relevant to insurance pricing.

[0044] 3. A machine learning model training module, which is used to train a prediction model based on the extracted group and individual vehicle features, which can predict the probability of insurance claims for the vehicle.

[0045] Specifically include: A feature selection unit, which is used to select the most influential features from the processed data; A model building unit, used to build a neural network model including an input layer, a hidden layer and an output layer; A training algorithm unit, used to train the neural network model using a back propagation algorithm and a gradient descent method; In this embodiment, the machine learning model training module uses deep learning technology, especially neural networks, to build an insurance pricing model. The steps of model training include: Feature selection: Use correlation analysis and feature importance assessment methods to select the most influential features from the processed data.

[0046] Model construction: Build a multi-layer perceptron (MLP) neural network, including an input layer, multiple hidden layers, and an output layer. The number of nodes in the input layer corresponds to the number of selected features, and the number of nodes in the hidden layer is determined according to the complexity of the problem.

[0047] Training algorithm: The neural network is trained using the back-propagation algorithm and the Adam optimizer. The loss function is defined as the mean squared error (MSE) to minimize the difference between the predicted risk and the actual risk.

[0048] Model validation: The generalization ability of the model is evaluated through cross-validation method to ensure that the model can also have good predictive performance on unknown data.

[0049] The trained model is used to predict the insurance risk of the vehicle in real time and adjust the insurance premium accordingly.

[0050] 4. Insurance pricing module: used to dynamically adjust the vehicle’s insurance premium based on the output of the prediction model.

[0051] A risk assessment unit, used to assess the insurance risk level of the vehicle; The rate adjustment unit is used to adjust the insurance premium rate based on risk assessment results and market data; In this embodiment, the insurance pricing system also includes a user interface for allowing a user to input vehicle information and view insurance pricing results and related risk assessment reports.

[0052] 5. Feedback learning module, used to continuously optimize the machine learning model based on insurance claims results.

[0053] The feedback learning module is used to continuously optimize the machine learning model and improve the accuracy of insurance pricing. The implementation steps of this module include: Data collection: Regularly collect data from insurance companies’ claims records, including claim amounts, claim frequency, and claim reasons.

[0054] Model evaluation: Analyze the differences between claims data and machine learning model predictions to evaluate model accuracy and bias.

[0055] Parameter adjustment: Based on the evaluation results, adjust the parameters of the machine learning model, such as weights and thresholds, to reduce the prediction error.

[0056] Model retraining: Retrain the model using the latest data to adapt to new data distribution and market changes.

[0057] In addition, the feedback learning module also includes a simulation system to test the effect of model adjustments before actual application to ensure that the adjusted model can improve the accuracy of predictions.

[0058] 6. Prediction simulation module: The prediction simulation module is used to simulate insurance risks under different driving behaviors and environmental conditions to evaluate the impact of different factors on insurance pricing. The implementation steps of this module include: Scenario definition: Define multiple driving scenarios, including different weather conditions, traffic congestion levels, driving behavior patterns, etc.

[0059] Data simulation: Use historical data and statistical models to generate simulation data to simulate the vehicle driving data under the above defined scenarios.

[0060] Risk simulation: Apply machine learning models to analyze simulated data and predict the risk of vehicle accidents in different scenarios.

[0061] Result analysis: Analyze the simulation results, identify which factors have a significant impact on insurance risks, and propose risk mitigation measures accordingly.

[0062] The predictive simulation module also includes a user interface that allows users to customize simulation scenarios and view risk prediction results under different conditions, thereby better understanding risk factors and making more informed insurance decisions.

[0063] 7.Data security module: In this embodiment, the design and implementation of the data security module aims to ensure the security and privacy of sensitive data generated by intelligent connected vehicles during the entire processing and storage process. The module includes the following key components: Data encryption: All data collected from vehicle sensors and systems is encrypted using the Advanced Encryption Standard (AES) before transmission. Symmetric key encryption algorithms are used to ensure data security during transmission; Anonymization: During the data processing stage, all personal identification information (such as vehicle identification number, driver identity information) is anonymized through a hash function to protect user privacy; Access control: Implement role-based access control (RBAC) policies to ensure that only authorized users can access specific data sets. Use multi-factor authentication mechanisms to enhance account security; Data auditing: Deploy a data auditing system to record all access and manipulation of data. These logs are reviewed regularly to monitor and prevent potential data abuse or security threats; Compliance Checking: The module includes a built-in compliance checker that automatically detects and ensures that all data processing activities comply with current data protection regulations and standards, such as the EU’s General Data Protection Regulation (GDPR).

[0064] 8: Intelligent recommendation module: The intelligent recommendation module uses machine learning algorithms to analyze users' behavior patterns and preferences to provide personalized insurance product and service recommendations. The following are the specific implementation steps of this module: User behavior analysis: Build user behavior profiles by analyzing users’ driving data (such as mileage, accident records, and driving time preferences) and feedback data (such as inquiries and purchase history for specific insurance products).

[0065] Preference identification: Use clustering algorithms (such as K-means or hierarchical clustering) to group users and identify different user groups and their preferences.

[0066] Product matching: Based on user preferences and behavior profiles, the intelligent recommendation system matches the most suitable insurance products. For example, for users who often drive long distances, higher levels of vehicle insurance and emergency rescue services are recommended.

[0067] Dynamic recommendation updates: The system continuously monitors users’ new behaviors and feedback, and dynamically adjusts the recommendation list to ensure the relevance and timeliness of the recommendations.

[0068] User interface interaction: Recommendation results are presented through a friendly user interface and customization options are provided, allowing users to adjust recommendation settings according to their needs.

[0069] 9. Real-time monitoring module: The real-time monitoring module is designed to track the vehicle's driving status and driving behavior in real time to adjust insurance rates and provide instant feedback. The implementation of this module includes: Real-time data streaming: Establish a real-time data streaming system to transmit vehicle driving data to the central server in real time through the vehicle communication system (such as 4G / 5G).

[0070] Real-time risk assessment: Develop a real-time risk assessment engine that analyzes real-time data (e.g., speed, acceleration, location) and calculates the current risk level.

[0071] Dynamic rate adjustment: Dynamically adjust insurance rates based on real-time risk assessment results. For example, automatically increase rates during high-risk periods (such as at night or in severe weather conditions).

[0072] Driving behavior feedback: Provide real-time feedback to the driver through the vehicle information system, such as safe driving tips or warnings, to encourage safe driving behavior.

[0073] System Integration: Integrate the real-time monitoring module with other systems (such as the accident prediction module and the intelligent recommendation module) to provide more comprehensive services, such as automatically recommending safe driving courses or insurance products when high-risk driving behaviors are detected.

[0074] This example demonstrates how to implement a comprehensive, dynamic, and user-friendly smart connected car insurance pricing system by combining advanced data processing, machine learning, and real-time monitoring technologies.

[0075] 10. Accident prediction module: The accident prediction module uses advanced machine learning technology to predict the possibility of a vehicle accident. The following are the specific implementation steps of this module: Data integration: Collect vehicle driving data, including speed, acceleration, braking frequency, historical accident records, traffic violation records, etc.

[0076] Feature engineering: Extract useful features from raw data, such as frequency of emergency braking, number of speeding, nighttime mileage, etc.

[0077] Model training and validation: Use historical data sets to train and validate the model to ensure the model's predictive accuracy and generalization ability.

[0078] Real-time prediction: Deploy the trained model to the vehicle’s real-time data processing system to analyze vehicle driving data in real time, predict accident risks, and issue warnings to the driver or vehicle management system in a timely manner.

[0079] 11. User behavior analysis module The user behavior analysis module analyzes users’ driving habits and preferences: Behavioral data collection: Collect users’ driving behavior data, such as driving speed, acceleration, braking frequency, etc.

[0080] Habit Identification: Use cluster analysis to identify different groups of driving habits.

[0081] Preference analysis: Analyze user preferences through their responses to different insurance products.

[0082] Personalized services: Provide personalized insurance services and product recommendations based on user behavior and preferences.

[0083] Feedback loop: Establish a user feedback mechanism to continuously optimize the accuracy of user behavior analysis.

[0084] Through the insurance pricing system disclosed in this embodiment, the system can collect risk data of the entire group of intelligent connected vehicles connected to the autonomous driving network in the area, and then add machine learning model training to perform a prediction model for group and individual vehicle feature training, which can predict the insurance claim probability of the vehicle and perform more accurate insurance pricing; At the same time, this system can dynamically adjust the vehicle's insurance premium rate and conduct dynamic pricing through the dynamic monitoring and networking functions of intelligent connected vehicles, avoiding problems such as inaccurate vehicle risk assessment and large pricing errors due to information lag, and is suitable for promotion and use.

[0085] The above is only for explaining the implementation mode of the present invention and is not intended to limit the present invention. For those skilled in the art, any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention without creative work should be included in the protection scope of the present invention.

Claims

1. A method for pricing autonomous driving insurance based on machine learning, characterized in that: The following steps are involved: Collect data from all intelligent connected vehicles connected to the autonomous driving network in the region, clean, denoise, and standardize the collected data, integrate individual vehicle data with regional traffic data, and build a large database of group connected vehicles; Divide smart connected vehicles into different regional groups based on the geographic location information in vehicle data; Analyze the environmental characteristics of each regional group, use spatial statistics and time series analysis methods to identify regional risk hotspots and change trends, and obtain regional risk coefficients; Analyze the communication data between vehicles in the regional group, build a risk model for vehicle interaction, and obtain the risk coefficient for vehicle interaction; The vehicle individual risk coefficient and driving behavior adjustment factor are obtained according to the smart networked car data, and a group networked car insurance pricing model is constructed by combining the regional risk coefficient and the vehicle interaction risk coefficient.

2. The method for pricing insurance for autonomous driving based on machine learning according to claim 1, characterized in that: The insurance premium pricing method of the group networked car insurance pricing model is: , where β is the weight coefficient of vehicle-to-vehicle interaction risk in the overall risk assessment, which is determined by the insurance company’s strategy and market data.

3. The method for pricing insurance for autonomous driving based on machine learning according to claim 1, characterized in that: The method for obtaining the regional risk coefficient is: Among them, j represents different areas, nj represents the number of groups in area j, the accident severity coefficient is based on the accident information entered into the autonomous driving network after the accident, and the traffic congestion impact factor is determined according to the real-time road condition information at the time of the accident.

4. The method for pricing insurance for autonomous driving based on machine learning as claimed in claim 3, characterized in that: After obtaining the regional risk coefficient, a machine learning model is established for the regional risk coefficient to perform dynamic analysis and prediction. The machine learning model is: , where RegionRiskMap t+1 represents the regional risk distribution map at the next moment, ConvLSTM(.) represents the convolutional long short-term memory network model, SpatialFeaturest and TemporalFeaturest represent spatial features and temporal features respectively.

5. The method for pricing insurance for autonomous driving based on machine learning according to claim 1, characterized in that: The method for obtaining the vehicle-to-vehicle interaction risk coefficient is: , where m represents different types of vehicle-to-vehicle interactions, and the interaction type risk coefficient, interaction frequency, and interaction distance influencing factors are determined based on the real-time data of smart connected cars during driving.

6. The method for pricing insurance for autonomous driving based on machine learning according to claim 5, characterized in that: After obtaining the inter-vehicle interaction risk coefficient, a machine learning model is established for the inter-vehicle interaction risk coefficient to perform dynamic analysis and prediction, and the machine learning model is: , where InteractionRisk m,uv represents the interaction risk between vehicles u and v under interaction type m, GNN(.) represents the graph neural network model, VehicleFeatures u / v represents the feature vector of vehicle u / v, EdgeFeatures uv represents the features of the interaction edge between vehicles u and v, and GraphStructure represents the structural information of the vehicle interaction network.

7. The method for pricing insurance for autonomous driving based on machine learning according to claim 1, characterized in that: The risk assessment method for constructing a group network automobile insurance pricing model is: ， Among them, OverallRiskScore represents the overall risk score of the group of connected cars, wi represents the weight of the i-th basic model in the ensemble learning, Model i (.) represents the i-th basic risk assessment model (such as XGBoost, Random Forest, etc.), RegionRisk, VehicleRisk, and InteractionRisk represent the assessment results of regional risk, individual vehicle risk, and interaction risk between vehicles, respectively.

8. The method for pricing autonomous driving insurance based on machine learning according to any one of claims 1 to 7, characterized in that: The group networked vehicle insurance pricing model is dynamically adjusted according to the intelligent networked vehicle data, generates a dynamic risk assessment model, and dynamically adjusts the insurance premium pricing according to the dynamic risk assessment model.

9. An autonomous driving insurance pricing system based on machine learning, characterized in that: Applying the insurance pricing method according to any one of claims 1 to 8, comprising: Data collection module: used to collect driving data of the entire group of intelligent connected vehicles connected to the autonomous driving network in the area; Group data processing module: used to clean, standardize and feature engineer the collected group data to extract features related to insurance pricing; Machine learning model training module: used to form a prediction model based on the extracted group and individual vehicle characteristics, which can predict the probability of insurance claims for vehicles; Insurance pricing module: used to dynamically adjust the vehicle's insurance premium based on the output of the prediction model; Feedback learning module: used to collect insurance claim results, integrate the insurance claim results with the group data and optimize the machine learning model.

10. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the insurance pricing method according to any one of claims 1 to 8 is implemented.

Images