Graph convolutional network approach applied to predict hourly bike-sharing demands considering spatial, temporal, and global effects

• DOI: 10.1371/journal.pone.0220782
• Publication Date: 2019-09-16T17:28:15Z
• Authors: Tae San Kim; Won Kyung Lee; So Young Sohn
• Keywords: Bike Sharing; Baseline (sea)

Solving the supply-demand imbalance is most crucial issue for stable implementation of a public bike-sharing system. This gap can be reduced by increasing accuracy demand prediction considering spatial and temporal properties bike demand. However, only few attempts have been made to account both features simultaneously. Therefore, we propose framework based on graph convolutional networks. Our reflects not dependencies among stations, but also various patterns over different periods. Additionally, consider influence global variables, such as weather weekday/weekend reflect non-station-level changes. We compare our other baseline models using data from Seoul's Results show that approach has better performance than existing models.