Research On Driver Style Identification And Speed Prediction Using Driving Monitoring Data

With the development of road transportation industry,the contradiction between road passenger service capacity and travel demand is gradually intensified,and road accidents happen frequently.As the driver of the vehicle,his unsafe behavior is the main cause of road accidents.Onboard the popularity of smart devices to facilitate the traffic data acquisition for a long period of time,through the mining traffic data and extract the valuable driving behavior characterization parameters,control the driver behavior,on the basis of the research on vehicle motion state,provide a reference for driver warning unsafe behavior,is of great significance to improve the level of road transportation safety.Based on the results of driving style identification,this paper starts from the traffic data mining and carries out predictive research on the vehicle speed,which is the most important vehicle motion parameter,in the hope of improving the speed prediction accuracy and providing applicability to different drivers.The main research contents are as follows:(1)Research on driving style identification method.The driving style characteristic parameters involved in the current research and their applicable scope are summarized.Aiming at the possible multicollinearity among multiple driving style characteristic parameters,a feature dimension reduction method based on principal component analysis is proposed.A driving style identification algorithm based on the improved binary Kmeans algorithm is proposed.Aiming at the problem that the initial clustering center in the bisecting K-means algorithm seriously affects the clustering result,i Forest algorithm is used to calculate the data outliers and the method that the average difference degree is used as the evaluation index is improved.In order to make a contrast with the driving style identification model based on the improved binary K-means algorithm,FCM algorithm and spectral clustering algorithm commonly used in the clustering algorithm were used to construct the driving style identification model respectively.(2)Study on speed prediction method considering driving style.In order to represent the changing trend of the speed,a fuzzy reasoning machine with the speed deviation and driving aggressiveness as input and the speed correction coefficient as output is developed by using the fuzzy control algorithm.Aiming at the problem of drivers’ behavior characteristics not taken into account in the existing research on speed prediction,a DSNAR speed prediction model was established based on NAR neural network based on the input of speed correction coefficient and historical speed information.(3)Example verification and analysis.The model verification experiment was carried out with the driving data of 30 drivers on the passenger line from Chongqing to Dazu in one quarter,and the Hadoop platform was built as a data processing and analysis tool.Driving style identification experiments show that the improved bisecting K-means driving style identification algorithm has a better driving data clustering results,the improved binary K-means driving style identification algorithm to 30 intercity passenger driver data gathered for the cautious type,ordinary type,radical type 3 class,through the analysis of the driving style class cluster characteristic parameter,prove that defined the driving style labels and safety characteristics show good consistency;Speed prediction experiments show that DSNAR speed prediction precision of the model is better than that of only considering the speed of sequence data HMM model,BP neural network model and the NAR neural network model,according to the different speed prediction under the prediction step length,according to the results of DSNAR speed prediction model within10 s step has higher prediction accuracy,alert to unsafe driving behavior has a certain reference value.