| In recent years,with the rapid development of urban computing,intelligent transportation system(ITS)has become a popular research field based on location service.As one of the important research directions,metro flow forecasting can not only provide important decision-making foundation and data support for the intelligent scheduling and planning of urban subway,but also meet the strong demand of travel users to avoid subway station congestion.Although the current urban metro generally owns the scheduling and management system,but still exposed the mismatch problem between supply and demand,resulting in safety hazards.If a metro forecasting system can be applied to predict the passenger flow at subway stations,it will greatly improve the operational efficiency of the subway network and meet the travel requirements of citizens.The application of deep learning technology has gradually risen in the academic field,the deep spatial-temporal network formed by the combination of spatiotemporal prediction theory,has strong characteristic performance ability to fit nonlinear comprehensive data,which can capture the spatiotemporal effect of traffic flow.It has been widely attempted in flow prediction algorithm at this stage,and produced a number of results.In this paper,a metro flow forecasting modeling scheme is proposed,based on the city subway swipe data,using a deep spatial-temporal network that integrates attention mechanism and multi-scale convolution,learns potential temporal and spatial relationships,adds the global context external information and multi-task learning strategy.The model can not only improve the accuracy,but also promote the stability and generalization ability of this incoming/outgoing passenger flow forecasting task of subway stations.This paper then realizes a metro flow forecasting system to solve the problem of intelligent operation.Adhering to the theory of software engineering,the procedure of system building had been strictly controlled in this paper,and eventually completed all the construction work of the platform with web side and server.First of all,the requirements analysis was carried out to determine system requirements.Secondly,the system was outlined,the overall architecture and technical selection were coordinated,several functional modules were divided.In addition,the system database design and interface design are conducted.Then the detailed design and implementation of the system was expounded,construction of the algorithm model and development of the project were explained in detail.Finally,the system had been fully tested around function and performance. |
