Feature Mining Of Passenger Point Based On Taxi Trajectory Data

Today,with the rapid development of economy,the city as a regional political center concentrates on the core functions of a region's economy,culture and commerce,and the density of population is increasing year by year.On the basis of the limited traditional public transport resources,the rapid increase in population size has brought challenges and pressures to the normal operation of urban transport.As an important part of public transportation,taxis are highly favored by residents for their efficiency,convenience and flexibility.With the rapid development of information technology,floating car technology has been widely applied to taxi in big cities.In the daily operation of taxis,a large amount of trajectory data and vehicle status related information will be generated.Mining and analyzing these data scientifically and reasonably is of great significance for the study of taxi waiting berths,management scheduling and commuter characteristics.This paper focuses on the mining and analysis of taxi passenger characteristics after thorough understanding of the current research status of taxi trajectory data mining.Because of the accuracy of GPS sensor and the interference of environmental factors,the original trajectory data contains a large amount of noise data,which is not suitable for direct further research,and needs to be preprocessed.First of all,in order to ensure that the trajectory density of the study area is not disturbed by special outliers,the municipal area is taken as the main research area through the latitude and longitude coordinate transboundary processing to remove the trajectory data which is not included in the designated area.Secondly,the redundant data which caused by special conditions such as slow driving,vehicle congestion,and equipment anomalies can be eliminated by using the Douglas-Peucker algorithm.Then,the trajectory trip point caused by the sensor's low accuracy,environmental disturbance,etc.is calculated and recognized,and the identified erroneous trajectory points are eliminated.Finally,passenger status data in the taxi trajectory data is extracted using the switching of the passenger status based on the status information of the vehicle in the taxi track,so as to pave the way for the next research work.Next,the taxi passenger number,the passenger peak period and the distribution characteristics of the operating vehicle can be obtained by carrying on the statistical analysis to the pre-processed data in this article.From the three indicators,it can be intuitively found that the activity of the residents on weekends and holidays is reduced compared with the working days,and the direct peak and indirect analysis are used to get the daily peak hours of taxis.Through careful comparison,it can be found that residents activity heat slightly decreased during the working day residents are compared on weekends.Then the DBSCAN algorithm is used to cluster the passenger-carrying data during the peak hours of passengers on the working day and is improved in the short-term time-span,after that the passenger-carrying data is analyzed according to different time windows and granularity.The experimental results show that the improved algorithm can better express the distribution characteristics of the passenger hotspots compared with the original algorithm,and has obvious advantages in the expression of the moving trend of the passengers' points with the time changing.Considering the use of more influencing factors to mine and predict the characteristics of taxi passengers is the focus of the next step.