Research On Passenger Driver Driving Behavior Analysis And Speed Forecasting Method

With the development of transportation industry,the contradiction between highway passenger service capacity and travel demand is gradually intensified,and road traffic accidents occur frequently.As the operator of the vehicle,the driver’s unsafe behavior is the main cause of road accidents.The popularity of on-board intelligent devices facilitates the acquisition of long-term driving data.By mining driving data and extracting valuable driving behavior characteristic parameters,drivers’ behavior habits can be mastered.Based on this,vehicle motion state prediction research can be carried out.This paper selects the driving data of passenger drivers provided by a passenger transportation company in Chongqing,and uses data mining technology to mine the characteristic parameters of driving behavior from the driver’s point of view.Based on the clustering method,and the speed prediction model is established on the basis of the driving behavior analysis.By predicting the future driving state of vehicles,the warning and prompt basis for drivers’ unsafe driving behaviors can be provided.It can also provide a basis for the traffic management department to carry out targeted safety education and training.The main research contents are as follows:(1)Selection of characteristic parameters for driving behavior analysis and fusion of classification attributes.The redundant or less influential attribute data are eliminated to reduce the difficulty of the analysis of passenger car travelling data.(2)Driving behavior analysis.Driving behavior analysis in this paper,a reference when driving skills and driving style two indicators,categorizing driving behavior in the form of clustering,an improved Quan Shang FCM driving clustering analysis algorithm,the algorithm for Quan Shang FCM algorithm can only classes for clustering of data within a cluster spacing smaller and cannot achieve all kinds of problem of the distance between the clusters,Modified parameters were introduced to improve the distance between clusters after clustering,and then the clustering effect was optimized.In order to compare with the improved weight entropy FCM driving behavior clustering algorithm with added correction coefficient,the clustering models of weight entropy FCM algorithm and k-means algorithm were selected.In addition,the classification results of driving behavior without considering weather and the classification results of driving behavior considering weather factors were compared,and the influence of weather factors on driving behavior was verified.(3)Establish the speed prediction model based on the driving behavior classification results of natural driving data.Considering the large amount of driving data obtained by passenger drivers based on the passenger company monitoring system in this paper,the speed prediction model of BP neural network with artificial intelligence characteristics is selected to predict the speed.In order to solve the problems that the speed prediction model of BP neural network can’t jump out easily in the local minimum state and the connection weight adjustment is slow in the flat area,a BP neural network speed prediction model is proposed,which combines the adaptive learning rate adjustment,the additional impulse term and the simulated annealing algorithm.(4)Verification of model experiment.With the improved Quan Shang FCM clustering algorithm driving behavior analysis without considering the natural driving data under the weather for clustering analysis,clustering after driving behavior can be divided into radical,general skilled,skilled composed type three categories of skilled,comparing the classification results under considering the weather factors,driving behavior classification analysis is verified by the weather factors,In addition,the clustering results of weight entropy FCM algorithm,kmeans algorithm and the improved weight entropy FCM driving behavior analysis clustering algorithm were compared,and the conclusion was drawn that the improved weight entropy FCM driving behavior analysis clustering algorithm had the least number of iterations,the largest distance between clusters and better clustering effect.Driving behavior classification result vector as a parameter to the adaptive adjustment,additional impulse and organic combination of simulated annealing algorithm is integrated in the speed of the improved BP neural network prediction model is trained and speed prediction,in addition to select speed of BP neural network forecasting model and speed prediction model based on gray sequence.By comparing the errors between the predicted speed values and the actual speed values,it is proved that the improved BP neural network model has a better prediction effect.