FlowCasting: A Dynamic Machine Learning based Dashboard for Bike-Sharing System Management

Prompted by increasing citizens’ demand, the rapid evolution of smart- and micro-mobility continues to shape the landscape of urban transportation services. In light of their practical benefits in terms of environmental sustainability, public health, and traffic congestion mitigation, smart cities manage mobility services by tracking user demand and service utilization over time. Leveraging this data is crucial for discerning current patterns and anticipating future trends, thus improving service provision. In this context, we propose a new interactive dashboard for the advanced analysis of spatio-temporal data acquired from bike-sharing systems. Our goal is to show on an interactive map the city areas with the highest current and future users’ demand and a simulation of the routes suitable for redistributing bikes across stations according to their predicted occupancy level. We leverage a clustering algorithm to identify the areas with currently highest bike demand and a forecasting approach to predict users’ demand trends. Thanks to multi-resolution time and path management, end-users can exploit the dashboard to support their decisions regarding resource shaping. We showcase the FlowCasting’s capabilities on a opensource dataset collecting BlueBikes data in Boston (U.S.). The online demo is available at the following link: https://flowcasting.streamlit.app/