Analysis Of Characteristics Of Residents' Travel Behavior Based On GPS Trajectory Data

With the increasing number of taxis in the city,taxis occupy an increasing proportion in people's daily travel mode.Due to the continuous development and popularization of GPS technology,many taxis are equipped with GPS system.These taxis can record the location,time,orientation and speed of the vehicle during the course of driving.Through processing and analyzing the data of trajectory,mining potential value has become a hot research field.Because of the increasing number of taxis,the volume of taxi trajectory data is constantly increasing,and the standalone calculation and processing model in the past is no longer competent.Therefore,the big data technology is used to mine the trajectory data,and the Spark platform has the ability to store and manipulate big data.Based on Spark platform,this paper uses Spark parallelizable K-Means algorithm to mine taxi trajectory data and then analyze Chengdu residents' travel behavior characteristics.The meaning of residents in the paper refers to all passengers based on taxis in Chengdu,not to those who live in a place in the traditional sense.The main work includes the following points:(1)Preprocessing taxi trajectory data.Due to man-made improper operation,machine error and accidental error,source trajectory data need to be pre-processed before being applied to trajectory data mining.In this paper,the preprocessing of trajectory data mainly includes four aspects: distortion data elimination,invalid field deletion,invalid period data deletion and map matching.(2)Analyzing the law of residents' travel time.Based on the Spark platform,the relative algorithms are used to calculate the total number of residents' day trips and the number of trips during each hour of the day.Contrasting the difference of travel amount on weekday and weekend and the travel amount in different time periods,and analyzing the causes.(3)Analyzing the raw of residents' travel distance.Spark platform is used to analyze the raw of residents' travel distance from two aspects.Firstly,the proportion of residents' travel from different distances is analyzed.Secondly,the average distance traveled by residents for each hour of the day is analyzed.(4)Extracting and analyzing the hot spots of residents' trips.Based on the Spark platform,the K-Means algorithm is used to cluster the residents' travel trajectory data,extract the travel hot spots of urban residents,analyze the number and distribution of hot spots,and then analyze the characteristics of residents' travel behavior.