Vehicle Driving Profile Analysis And Risk Research Based On Data Of Internet Of Vehicle

The continuous development of China's economy and the steady improvement of people's living standards lead to an increase in the number of vehicles,which brings new problems such as traffic safety,road congestion and environmental pollution,and also brought new challenges to the auto insurance industry and the traffic supervision department.The Internet of vehicle technology based on the concept of Internet of Things and the data mining theory based on statistical method provide new ideas for solving these problems and challenges.This paper uses the data of company N,a famous domestic IoV data platform,to build a vehicle driving profile system and a risk analysis model.Firstly,the data is preprocessed according to the data mining theory.Then,from the various scenarios in the actual driving of vehicles,a series of scene variables are extracted and constructed,the descriptive statistical analysis is also carried out on the variable set.Based on the scene variables and descriptive statistics,several of variable extraction methods are compared,and then the sparse principal component analysis and K-means clustering analysis are used to design a multi-level and multi-dimensional vehicle driving profile SVT system.In the vehicle risk research part,this paper notices the characteristics of risk data,establishes a variable selection-rebalance-classification algorithm framework,and compares various methods.The SVT vehicle profile system proposed in this paper can realize the hierarchical characterization of the scene variable dimension,vehicle dimension and time dimension of the vehicle.The LASSO+oversampling+random forest algorithm under the framework of variable selection-rebalance-classification algorithm also achieves high risk prediction accuracy.The research results have high commercial application value for the innovation of IoV related products and UBI auto insurance pricing.