Research On Behavioral Patterns Of Taxi Passengers And Drivers

With the rapid advancements in science and technology,along with socio-economic development,on the one hand,the process of urban modernization is accelerating;on the other hand,with the development of city,the phenomenon of city traffic congestion is increasingly common.As the vital component of urban transportation,the research of behavior patterns of traffic participants in intra urban area is of great importance in many aspects,such as deeply understanding human travel behavior and relieve the phenomenon of traffic congestion.In the study of human mobility and activities,vehicle equipped with GPS locator is a very important sensing device.As a major part of urban traffic,taxi equipped with a GPS device meets a large number of traffic travel needs of people in the intra urban area.At the same time,taxi trajectory data generated by GPS device contains various laws,attracting the extensive concern of many scholars.Based on the previous study of group behavior patterns generated by taxi data,this paper will dig behavior patterns of traffic participants from two aspects as follows: the trip laws of taxi passengers in the intra urban area generated by taxi data and the prediction of taxi drivers’ passenger-seeking strategy.The main contents of this paper are divided into two aspects:(1)Based on the mining of the taxi travel data,three statistical indicators,which reflect travel patterns of taxi passengers in the intra urban area generated by taxi,are discussed: trip displacements and duration time that directly reflect taxi passengers’ travel pattern,and vacant time that indirectly reflect traffic demand of taxi passengers.Through the comparison with the results concluded by other domestic and foreign scholars,as well as our analysis of five taxi datasets from three cities,our paper finds that the lognormal distribution with a power-law tail,rather than power-law distribution nor the exponential distribution,can best approximate both the displacements and the duration time of taxi trips.In other words,for the short-distance travel in the intra urban area,the trip displacement follows the lognormal distribution;while for the long-distance travel,the power-law distribution fit best.The same law is also found in the duration time and vacant time,in all the examined cities.The universality of the scaling laws of taxi passengers’ mobility is subsequently discussed,in view of the analysis of the data.In addition,we find that the patterns of taxi passengers are stable,not changing with time and region.(2)Through the extraction of taxi trajectories data,we will predict the passenger-seeking strategy of taxi drivers.Firstly,we will find the top pick-up and drop-off hotspots from all the taxi drivers’ trajectories based on the method of grid partition.Secondly,combining the hotspots with the time slots,we will identify the function of specific region.Finally,we will convert the passenger-seeking strategy of taxicab drivers into feature quadruplet,namely Time-Region-Speed-Behavior.Based on the data and selected features,we propose a random forest model established by our feature quadruplet.The classifier can achieve the prediction accuracy of 95.01% on the new dataset,indicating that the proposed model based on our feature quadruplet can describe the passenger-seeking strategy of taxi drivers well,and the selected features can be used as the indicator to guide the passenger-seeking strategy of taxi drivers.