Data-driven Taxi OD Demand Forecasting And Visual Analysis System

With the advancement of urbanization,the number of motor vehicles on the road has been increasing sharply,which results in traffic congestion and brings new challenges to traffic management.Nowadays,it is urgent to alleviate the road pressure and optimize the travel experience of urban residents.Short-term OD prediction is of great significance in the field of intelligent transportation.This thesis aims to predict the traffic flow at the beginning and end of a certain period in the grid under the spatial scale division.The core of accurate prediction is how to fully mine the temporal and spatial characteristics and periodic characteristics of traffic data.Due to its limitations,the traditional prediction methods are difficult to ensure the quality of prediction results,and the existing studies rarely integrate external factors such as weather and road characteristics into the category of OD prediction.In this context,this thesis proposes the K-STACGN model,which is based on the GPS data and order data of taxis in Xiamen under the spatial grid division,comprehensively considers the spatio-temporal periodic dependence and spatio-temporal similarity characteristics of traffic data,and integrates external factors such as weather and point of interest(POI)information at the urban architectural level into the OD flow prediction in the future.At the same time,the attention mechanism is used to calculate the spatial weight of different time or feature map in the model to improve the prediction accuracy.In this thesis,different grids are selected for experiments.Through the experimental comparison with the classical neural network model,it is found that the combined model proposed in this thesis can improve the prediction accuracy to a certain extent.Commuting is an important source of OD traffic and the main travel demand of urban residents.Effective analysis of the passengers’ travel behavior and influencing factors of commuting activities under the commuting scenario can help to promote commuting activities,optimize the urban spatial structure and better meet the travel needs of commuters.In view of the lack of detailed spatial scale of commuting factor analysis and ignoring the characteristics of urban building facilities in OD grids,this thesis presents a mathematical regression model of commuting flow incorporating large multi-source space-time data.Under the partition of fine-grained spatial grids,traffic flow in Xiamen commuting scene is modeled mathematically,and explanatory variables are analyzed quantitatively.Commuting traffic data comes from taxi order data of Xiamen City,which reflects the travel characteristics of passengers more truthfully than OD forecast results.In addition,with the increasing amount of passenger flow,some taxis are in an "unconventional" operation state,such as carrying passengers across a long area and exceeding the regular operation time of individuals.Such kind of taxis are defined as "abnormal individuals".In order to better realize the scheduling and control of this kind of taxi,this thesis makes a clustering and visual analysis of this kind of "abnormal" travel group.On this basis,a visual analysis system is built to complete a variety of data display effects.