An Approach to Analyze Commuters’ Transfer Characteristics on the Public Transportation Network in an Incomplete Data Environment

This paper analyzed commuters’ transfer characteristics on public transportation networks in an incomplete data environment where only smartcard data could be accessed. A frequent-patterns-mining method was proposed, and a case study was given in Shenzhen, China. Specifically, commuters were firstly identified by using smartcard data based on their bus-riding regularity on workdays. Then, frequent correlations between bus routes/metro lines used in long-term rush hours were measured for each commuter. A two-level minimum threshold framework was proposed to evaluate the correlations, and then determine whether commuting trips containing transfers. The results indicated that non-transfer commuting trips accounted for nearly 72% of the total, among which about 87% were taking buses. Moreover, 14% of identified commuters were having commuting trips through one transfer, where interchanges within the same transportation mode were the majority, such as transfers from bus to bus, and transfers between metro lines. It reflected that the efficiency and convenience of transfers from bus to metro needed to be improved.