The Research And Application Of Traffic Bayonet Data Mining Technology Based On Hadoop

As society advances,people's quality of life is improving,the number of retained vehicles has been showing rapid growth.For one thing it makes people's lives more convenient and brings more convenience for their going out,for another it also leads to a series of traffic problems,such as traffic jams,traffic accidents and air pollution as well.Now the intelligent transportation system is the key to solving the traffic problems.As one of the important devices,intelligent monitoring and recording system(short for traffic bayonet system)has played an increasingly important role in urban traffic management.However,with the rapid growth of the number of bayonet system and vehicles,the bayonet information of car had a rapid expansion.Although there is a wealth of bayonet data of the car,appropriate methods and techniques to mine these massive traffic data behind the hidden valuable information is still lack.For this reason,through characteristics analysis of city bayonet data and literature study,the paper studies the massive bayonet data,which acts as its research object,in "lost driving" excavation and traffic flow forecasting based on the popular data processing technology Hadoop platform and combined with related data mining algorithms.It has important sense to alleviate the current more serious problem of traffic safety and traffic jams.This paper presents the research of lost driving based on massive historical bayonet data,main include the acquisition of suspect information,the analysis of vehicle trajectory aggregation based on spatiotemporal neighborhood and the prediction of the space-time trajectory of lost vehicles,and then presents the distributed design of the algorithm.It's to form a complete set of scheme from discovery and investigation of the "lost driving" illegal actions.In this paper,By analyzing the temporal and spatial association characteristics of traffic flow,a forecasting model of traffic flow based on temporal and spatial feature fusion isproposed.Given the K-Nearest Neighbor algorithm used time series traffic flow forecast model based on Hadoop and local linear embedding algorithm used traffic flow prediction model based on Hadoop for traffic flows of adjacent upstream streets,with the aid of a weighted combination of prediction model to determine the combination,the prediction accuracy and running time will be improved.At last,the distributed algorithms of two mining applications are transplanted to the Hadoop platform,and the feasibility and validity of the designed algorithm are validated by testing the real bayonet data.