Demand and Trip Prediction in Bike Share Systems

Bike Share systems are becoming increasingly popular in urban areas. With growing membership and expansion of service comes many operational challenges. A major challenge in their operations is the unbalanced demand and supply at bike stations as a function of time. Figure 1 shows number of bike trips in Jan 2017, aggregated into time intervals of 30 minutes according to start time, and summed across all days. We see that there is a clear temporal dependence of bike demand. Similarly, work districts have a higher demand during evening rush hours whereas residential areas have a higher demand during mroning rush hours. Most bike share systems employ active rebalancing to ease the pressure at peak times. This means transporting a certain number of bikes from inactive stations to more active stations, or between stations and storage, in order to maximize the usage of each bike and ease supply and demand inbalance problems across bike stations at di erent times. A quantitative, predictive model for the demand and supply would help operators plan bike transports more e ciently. This project aims to build such a model for bike arrivals at stations within one-hour time intervals, as a function of the following parameters: