Assessing and Predicting Mobility Improvement of Integrating Bike-Sharing into Multimodal Public Transport Systems

Abstract

New shared mobility services have become increasingly common in many cities and shown potential to address urban transportation challenges. This study aims to analyze the mobility performance of integrating bike-sharing into multimodal transport systems and develop a machine learning model to predict the performance of intermodal trips with bike-sharing compared with those without bike-sharing for a given trip using transit smart card data and bike-sharing GPS data from the city of Seoul. The results suggest that using bike-sharing in the intermodal trips where it performs better than buses could enhance the mobility performance by providing up to 34% savings in travel time per trip compared with the scenarios in which bus is used exclusively for the trips and up to 33% savings when bike-sharing trips are used exclusively. The results of the machine learning models suggest that the random forest classifier outperformed three other classifiers with an accuracy of 90% in predicting the performance of bike-sharing and intermodal transit trips. Further analysis and applications of the mobility performance of bike-sharing in Seoul are presented and discussed.