The Bus Travel Time Prediction Method Of Urban Road Base On AVL Data

With the continuous development of city construction,the urban road traffic environment is increasingly complex and changeable.When buses are running on urban roads,they will be disturbed by traffic signals,social vehicles,pedestrians and other factors,resulting in the instability of bus running state.Especially in developing countries,due to the excessive number of vehicles,the diversification of transportation modes and the serious lack of lane discipline,this randomness is further expanded,which greatly reduces the prediction accuracy of bus travel time.Therefore,this paper aims to analyze the characteristics of bus operation in China,explore the factors affecting bus travel time,divide the urban road traffic status,develop and verify the prediction algorithm of bus travel time,finally achieve the goal of improving the quality of bus service and increasing the attractiveness of bus.Firstly,the paper makes clear the importance and necessity of the research on the prediction method of bus travel time,summarizes and analyzes the research on the prediction of bus travel time by domestic inland and foreign scholars in recent years,summarizes the worthy of reference of this research,and then comb and determines the research content and research ideas of the paper.Secondly,the paper analyzes the structural characteristics of the collected Automatic Vehicle Location data,independently designs a data preprocessing algorithm for its existing data quality problems,and use Matlab2014 b software to write an algorithm program to extract the data information needed for the experiment.The paper analyzes the time-varying and fluctuating characteristics of bus travel time according to the driving rules of buses in China and the current situation of urban road status in China.The variation trend of bus travel time in weekend and weekday was statistically analyzed,and the variation rule of bus travel time in time and space was explored.By analyzing and collating different wave characteristic analysis methods and considering the quantitative index of improved wave characteristic of occurrence interval,the paper aims to analyze the temporal and spatial wave rule of bus travel time.Thirdly,the paper based on the Clustering by fast search and find of density peaks,from the perspective of the volatility of bus travel time,according to the traffic environment divide the urban road traffic status in time and space of the bus.the division and the result was compared with the experimental results of other clustering algorithm analysis,to verify the effectiveness of the proposed algorithm,and finally get high volatility and low volatility of volatility in three different fluctuation levels of road traffic statefourthly,the paper analyses and lists the factors influencing the characteristics of the bus travel time prediction model,and select Embedding method as the eigenvector prediction models feature selection.The paper select the Deep learning Neural network(DNN)as a forecast method,and combined with the road traffic state with different levels of volatility,designed the bus travel time prediction model based on CFDP-DNN framework,and through the Tensoflow Deep learning tools to complete the construction of the experiment.By comparing and analyzing the prediction results with other prediction methods,it is found that the prediction model method based on CFDP-DNN proposed in the paper has the highest prediction accuracy and reliability.Finally,the paper summarizes the general research content of the paper,puts forward the shortcomings and shortcomings of the current research in the paper,points out the main problems in the paper,and gives the future work plan and the next step of the research program.