Short-Term Forecast of Bicycle Usage in Bike Sharing Systems: A Spatial-Temporal Memory Network

Bike-sharing systems have made notable contributions to cities by providing green and sustainable mobility service users. Over the years, many studies been conducted understand or anticipate usage of these systems, with hope inform their future developments. One important task is accurately predict patterns systems. Although deep learning algorithms developed in recent years support travel demand forecast, they mainly used traffic volume speed on roadways. Few applied them bike-sharing Moreover, usually focus one single dataset study area. The effectiveness robustness prediction are not systematically evaluated. In this study, we propose a Spatial-Temporal Memory Network (STMN) short-term bicycles framework employs Convolutional Long Short-Term models feature engineering technique capture spatial-temporal dependencies historical data for task. Four testing sites evaluate model. These four include two station-based (Chicago New York) dockless (Singapore Taipei City). By assessing STMN several baseline models, find that achieves best overall performance all cities. model also superior urban areas varying levels bicycle during peak periods when high. findings suggest reliability predicting different types