Road Safety Behavior Analysis And Forecast Based On Traffic Big Data

The advancement of existing positioning and communication technologies can accurately record the time and space data of the vehicle and store it in the transportation platform.These traffic data often hide important information,use clustering technology to process time-space trajectory data,obtain trajectory classification and abnormal trajectory without supervision,and establish an effective road traffic safety prediction model for accident data.Provide important decision-making basis for vehicle supervision and management of the transportation management department,so as to achieve the role of auxiliary decision-making.This thesis is aimed at the audit management needs of transportation,and studies the analysis and prediction methods of road safety behavior.The main research work includes the following three aspects:(1)For distributed traffic data storage systems,fast-growing traffic data may cause system obstacles to block or data loss,and the original traffic data storage system cluster has been configured with non-object-oriented storage data types,etc.The transparent access method from data storage system to object storage cluster realizes the fusion of big data file system and object storage service,and improves the efficiency and reliability of traffic big data mining.(2)Based on the spatio-temporal trajectory data of the vehicle,considering the trajectory characteristics,the similarity trajectory clustering and anomaly detection methods with spatio-temporal features are studied.An MBR-D graph sequence similarity measure algorithm is proposed.The minimum circumscribed rectangle of the sub-track is divided by the MBR algorithm.The minimum similarity distance matrix of the time series continuum value of the sub-track is calculated by the DTW algorithm as the basis of the trajectory clustering.The K-Means clustering algorithm is used to divide the similar trajectories into multiple clusters,and then use the WCSS error evaluation criteria to find the appropriate trajectory clustering results,and finally apply it to the identification of abnormal trajectories.(3)Study the method of establishing road traffic safety prediction model,apply machine learning algorithm to accident safety level prediction,select characteristic factors that have important influence on traffic accidents,and establish three road traffic safety prediction models based on historical traffic data.The prediction accuracy of various models is analyzed,and the importance of each feature factor on traffic safety is given.