Construction Of Driving Style Recognition Model Based On Multi-feature Parameter

With the continuous improvement of the living standard,the number of cars in our country is increasing,which leads to the serious problems of road traffic safety and fuel consumption.The driving style of drivers is closely related to road traffic safety and automobile fuel economy.Therefore,based on the driver's operation information and vehicle status information collected by the driving simulator.Using multiple feature parameters as an indicator to evaluate driving style attribute.The driving style recognition model with good recognition ability is constructed in this paper.The main work of this paper is as follows:(1)Driving experiment design and data acquisition: using the driving simulator to design the experimental road model,20 drivers are recruited to complete the simulated driving task for collecting the driving operation data and the vehicle dynamic signal during the driving proc ess.The data sets of 49 dimensions were constructed by characteristic processing.The self-rating and professional scores of aggressive were carried out by the way of questionnaire,and the scores were collected statistically.The drivers were divided into clam type,normal type and aggressive type by k-means clustering method for getting the label data.(2)Establishing the driving style recognition model.Using support vector machine(SVM)and back propagation neural network(BPNN)to construct the supervised driving style recognition models,and the ability of the two models to recognize driving style is compared.The results show that the driving style recognition accuracy of SVM and BP neural networks is 93.0348% and 92.0398%,respectively.(3)Optimize the driving style recognition Model.Analyze the defects of the multi-parameter data set and its influence on the construction of the driving style recognition mode.In this paper,the principal component analysis(PCA)is used to extract the features of the multi-feature parameter data set,the first eight principal components are used as the feature input of the driving style recognition model,and verify the rationality based on the SVM model.Generating self-training SVM semi-supervised model and inductive Multi-Label Classification with Unlabeled data.The comparative study shows that iMLCU shows the best driving style recognition ability among the three models,and the fewer marking samples,the more obvious the advantages,so iMLCU is selected as the optimized model.