Construction And Application Of Safety Driving Evaluation Model In Fleet Management Platform

With the development of automobile industry at home and abroad and the improvement of people’s living standards,China is gradually becoming a big automobile country.While promoting economic and social development and improving people’ s lives,the automobile industry also brings traffic safety problems that cannot be ignored.In this dissertation,a driving behavior research method is proposed to solve the problems of noise,abnormal data sensitive and unquantifiable driving safety assessment in the existing research process of dangerous driving behavior recognition.Based on the real-time vehicle operating data collected by the on-board OBD equipment,this method combines relevant machine learning algorithms to construct a dangerous driving behavior recognition model and a driving safety evaluation model.Finally,it proposes a complete driving safety evaluation scheme.The main contents of the dissertation are as follows:(1)Data collection and preprocessing.Firstly,the working principle of vehicle data acquisition equipment is discussed.Secondly,the data flow direction of the acquisition platform is explained,and then the sample properties of the model data set are explained.Finally,the original driving data are preprocessed by Kalman filtering algorithm.(2)Construction of dangerous driving behavior recognition model.The imbalanced data classification algorithm based on ensemble learning(RS-SVM)is used to construct three kinds of dangerous driving behavior recognition models,including rapid acceleration,rapid deceleration and sharp turn.The parameters of RS-SVM algorithm are optimized by cross validation,so as to improve the accuracy of dangerous driving behavior recognition.Finally,the experimental results are compared with other algorithms.(3)Construct driving safety evaluation model.Based on the acceleration,turning,deceleration and rollover model,the driving safety evaluation index is established.The weight of the index is determined by principal component analysis and the driving scores of different drivers are obtained.The K-means algorithm is used to cluster the driving style of the driver,and finally the comprehensive evaluation of the driver ’s driving level is realized.(4)Design and implement a fleet management cloud platform.A Web-based fleet management cloud platform is designed and implemented using Java SSM framework.The platform integrates four functional modules: vehicle management,driver management,safety management and personal center.The experimental results show that the accuracy of dangerous driving behavior recognition model reaches 98.596%.The establishment of a driving safety evaluation model is beneficial to reduce the incidence of traffic accidents,improve the efficiency of fleet management,and at the same time provide a theoretical basis for standardizing driver’s driving behavior.