Travel Demand Prediction with Application to Commuter Demand Estimation on Urban Railways

Yohei Kodama, Yuki Akeyama, Yusuke Miyazaki, Koh Takeuchi

Abstract: Travel demand forecasting is a vital problem in the development of smart cities, infrastructure planning, and transportation management. The advent of contactless smart card systems has enabled the collection of data regarding daily transit and purchasing activities, providing a rich source of insights into citizen behavior. In this paper, we introduce a new problem of predicting changes in travel demand resulting from the installation of a new facility while preserving privacy. To address this problem, we propose a simple but effective supervised learning method that can capture the relationships between residential areas and existing facility locations, and exploit spatial features to forecast future demand in response to a new facility location. As a workable example, we employ real-world data to predict the future travel demand triggered by the installation of a new station in a railway system. Through extensive experiments, we demonstrate that our method improves the prediction accuracy.