Bike Sharing Traffic Pattern Prediction from Urban Environment Data for Automated Station Planning

Abstract

The planning process for bike sharing systems is often complex, involving multiple stakeholders and several considerations: finding hotspots in the potential demand, and serving that demand effectively with limited expenses, is a complex problem. In this work, we consider urban environment data from multiple sources to design a simple and explainable prediction model for bike sharing activity and traffic patterns, building an automated design pipeline to place stations in an area. A use case in New York City shows that our system can effectively plan a bike sharing system expansion, or even an entirely new system in a city, providing a valuable first step for the planning process and allowing system designers to identify gaps in existing systems and the locations of potential demand hotspots.