Clustering Subway Station Arrival Patterns Using Weighted Dynamic Time Warping

摘要

To better plan and schedule public transportation resources, it is crucial to understand the travel demand from any location at any time. In this article, we focus on analyzing the demand patterns for subway stations based on the tap in data at each station entrance. It has been reported that accurately predicting the arrival rates can help improve the travel experience, and prevent over-crowding in train carriages or platforms. We proposed a weighted dynamic time warping approach (WDTW) to adaptively cluster similar patterns from multiple stations. These similarities can be exploited in improving the prediction performance because spatial temporal information is better utilized. We demonstrated our approach and its effectiveness through a real data example.