An applicable method of real time passenger flow prediction.

Advanced Public Transportation System (APTS) is a main research field of Intelligent Transportation Systems. APTS is accredited one of the optimal approaches to solve the urban traffic problems. Based on the research of some critical theories such as transit network assignment and real time dispatching, APTS applies synthetically the advanced technologies such as information, data communication, electronics, global positioning system (GPS), geographic information system (GIS) and computer science to the public transportation system. APTS can also adjust the headway, monitor the transit vehicles, arrange the drivers and conductors, work out the schedules, provide the pre-trip, on-trip information to the travelers according to the real time information. By means of APTS, the level of service (LOS) will certainly be improved, passenger flow will be attracted to transit, economic benefit will be increased, and transit operation will step into the information and intelligent era. In China, more and more local governments and transit companies have realized the significance of the implementation of APTS. Many large cities are trying to deploy APTS. For example, in Beijing, Shanghai, Hangzhou, Dalian, some transit lines have deployed Public Transportation Intelligent Dispatching System (PTIDS), which is a kernel subsystem of APTS. With the implementation of PTIDS, a very important problem, real-time passenger flow prediction, become more and more outstanding. As we know, real-time passenger flow has relation to the determination of headway and dispatching mode. Hence, it is a critical theory to make sure the effective implementation of PTIDS. However, it is still not well solved, which has become an obstacle. Up to now, the headway and dispatching mode can only be made certain based on the experiences of the dispatcher in the present dispatching systems. Aiming at the research actuality, this paper presents an applicable method to predict passenger flow in real-time based on clustering analysis and stepwise regression analysis.