Dynamic Public Transit Origin-Destination Estimation Based On Data Mining

With the integrated applications of modern communication technology, network technology, and Global Position System (GPS) positioning technology in the public transport system, the research and development on the Advanced Public Transportation System (APTS) in China has achieved visible progress. As the key support data of APTS, the acquisition of dynamic transit Origin-Destination (OD) data is an important part of research in APTS. The promotion and application of the bus Integrated Circuit (IC) card payment system provides a new method and means for obtaining the dynamic transit passenger data. Through the method of data mining, we can get the dynamic transit passenger data needed for the real-time scheduling and operational management in APTS. This thesis bases the data mining technique to analyze and deal with the dynamic bus data to obtain real-time dynamic transit passenger OD data, which provides a scientific decision-making support for the public transportation management and public traveling.First, the thesis gives a comprehensive synthesis of state-of-the-art in data mining and passenger OD derivation, summarizes shortcomings of existing studies, and analyzes the reliability and practicality of the bus IC card data.Bus IC card data mining system is the basic platform for obtaining the transit passenger OD. After introducing the acquisition methods of the dynamic bus data, the thesis uses bus IC card data and GPS positioning data as the data sources to establish the dynamic bus data warehouse and proposes four-level structural data mining system application platform. Then, based on SPSS Clementine basic platform, it establishes the suitable public transportation IC card data mining system application platform, which includes data warehouse module, data mining engine module, data mining module, and graphical user module.The information of the boarding, alighting and transfer points of the bus IC card passengers is the key for getting the dynamic transit passenger OD. The thesis proposes a data mining method to estimate and obtain the above three parts of IC card information. In the acquisition of the boarding point, the thesis carries out the cluster analysis for the swiping time of the bus IC cards, gets the bus arrival time using GPS positioning data, and then establishes the time matching model to access passengers' boarding point information. In the acquisition of the alighting point, it proposes a method considering the travel chain and the passenger travel distance distribution, In estimating the transfer point, it uses the method based on the swiping time difference information of the bus IC card.Finally, based on the proposed method, the thesis carries out the case study using the bus IC card data in Beijing. It validates the algorithm by comparing the estimated data with the actual data; and then analyzes the dynamic transit trip characteristics of residents in Beijing based on the bus passenger OD.