Multi-Feature Passenger Flow Forecasting Based On Taxi GPS Trajectory Data

The city’s transportation system is the lifeblood of a city or even a country’s economy,culture and other fields continue to develop.In recent years,China’s urbanization has developed rapidly.As the population and the urban scale grow,residents’ demands for travel are also rising,the taxi is the one of the main transport for residents,taxi service is also the significant part of urban transport system,but at present,the taxi market is facing a very serious contradiction between supply and demand,through the taxi passenger flow forecasting,guidance directions for taxi market deployment can be provided.And the demands for people also can be meet greatly.For the construction of transport system to contribute the power.In this thesis,the use of statistical analysis,machine learning relevant technology and the real city taxi GPS trajectory data,to forecast the taxi passengers of the city.Provide the reliable data support to solve the traffic problem.Firstly,based on the pretreatment of the GPS trajectory data,the trajectory information of the taxi operation is extracted,the position of the passenger is determined,and the urban area is divided into the urban traffic small areas and clustered based on the grid division technology.Then,basing on the analysis of the travel law of the residents,the identification of the travel characteristics is carried out,not only the general characteristics are identified,but also the potential characteristics of moving are identified,and the concept of the active factor is put forward for the first time.The Finally,the model is established,which is the multi-feature passenger flow forecasting based on taxi GPS trajectory data,and the deep neural network algorithm is used to enhance the expression ability of the model,so as to predict the taxi passengers flow.In this thesis,the passenger flow of different traffic small areas in Shanghai central area is forecasted by different time slices,and the prediction results and the influence of potential characteristics and active factor on passenger flow forecasting results are analyzed.Finally,a quantitative analysis is made on the actual traffic volume of each day and the forecast value.It is found that the accuracy rate of regional taxi passenger flow forecast based on this method is as high as 96%,which proves that the effectiveness and superiority of our proposed method.