Cluster Analysis And Pattern Recognition Of Driver Braking Behavior Based On Truck Operation Data

In recent years,the number of trucks in China is increasing year by year.However,the proportion of road traffic accidents involving trucks is relatively high.Due to the technical parameters and loading of trucks,traffic accidents involved in trucks often cause great economic losses and casualties.Among the causes of accidents,improper operation by truck drivers accounts for the highest proportion.Therefore,it is of great significance to analyze the behavior of truck drivers for improving truck traffic safety.In order to mine the dynamic characteristics and hidden risk information of the trucks from the massive truck operation data,the thesis takes the truck driver’s braking behavior as the research object,and carries out the characteristic analysis of the truck operation data.A method suitable for cluster analysis of truck driver’s braking behavior is determined.A truck driver braking pattern recognition model based on an improved Topic model is proposed.The distribution of the truck driver braking pattern is excavated,and the type of braking behavior and braking pattern of truck driver are obtained.Firstly,based on the truck operation data after cleaning and noise reduction,the statistical analysis method is used to analyze the distribution,dispersion and correlation characteristics of truck operation data.Furthermore,the reconstructed phase space method is used to analyze the characteristics of the operating parameters of the truck.Taking the speed parameter as an example,it is verified that there is chaos in the time series of the operating parameters of the truck.Secondly,the segment data of the truck driver’s braking behavior is intercepted,and the parameter characteristics of the truck driver braking behavior are analyzed.Twenty-five characteristic parameters are calculated,such as the braking behavior duration,mean value,maximum value and variance of speed,longitudinal and lateral acceleration.In order to weaken the correlation between parameter terms,the parameters are reduced in dimension through factor analysis.The seven main correlation factors obtained are used as input parameters for the cluster analysis of braking behavior.The Fuzzy C-Means clustering algorithm and the Multi Kernel Fuzzy C-Means clustering algorithm are respectively used for the cluster analysis of braking behavior.According to the evaluation scores of the clustering results,9categories of truck driver’s braking behavior were determined.And it is proved that the Multi Kernel Fuzzy C-Means clustering algorithm is more suitable for the cluster analysis of braking behavior.In addition,the type of truck driver’s braking behavior is defined according to the factor index.Finally,the Word Vector model is used to construct the braking behavior dictionary.The relationship among braking data,braking behavior and braking pattern is established.In order to improve the accuracy of model recognition,the word sequence information of braking behavior is introduced to improve the Topic model.Furthermore,the traditional Topic model and the improved Topic model are respectively used to recognize braking patterns on the braking data.Combined with the subject consistency index,three braking patterns are determined.In addition,the applicability and superiority of the improved Topic model are verified by comparing the correlation coefficient between the reconstructed braking data of the training results and the original braking data.Then,according to the distribution of different braking behavior words in the braking pattern,three types of braking patterns are discriminated and defined.The accuracy of the proposed truck driver’s braking pattern recognition model is verified.The research process and conclusion of the thesis not only show the potential regularity in the truck operation data,but also prove the feasibility and effectiveness of the proposed analysis method and identification model.The research methods and conclusions can provide reference for truck drivers’ traffic safety evaluation,risk early warning,safety education and training,etc.