RESEARCH ON SMART CARD DATA MINING FOR MULTI-MODAL PUBLIC TRANSIT

Abstract

Taking advantages of Singapore’s distance-based AFC system, the smart card data records both passengers’ boarding and alighting activities and more importantly integrates both bus and metro trips in a same data frame. This provides a great opportunity to unravel some hidden mechanism of multi-modal public transit. This thesis is devoted to delivering deeper insights into multi-modal public transit from a systematical perspective and provide some intelligent tools for relevant agencies to enhance the service quality and reliability of the public transit. The original contribution of this thesis is two-fold: (1) Innovative perspective: This research systematically sheds new light on various unexplored aspects in multi-modal public transit, especially the transfer-related issues have been extensively investigated. (2) Advanced techniques: A group of novel prediction frameworks are proposed in accordance to different passenger flow prediction tasks. The advanced deep learning techniques empower the presented frameworks with superior accuracy, practicality and scalability.