A decision model for bike sharing based on big data analysis

Abstract

This paper aims to design an effective scheduling method for bike sharing system that minimizes the supply-demand gap in different areas. Based on the actual data on bike sharing in a city, the spatiotemporal features of bike sharing demand were analyzed through the prediction by a backpropagation neural network (BPNN), whose structure was determined through linear regression and other techniques. According to the predicted demands, multiple assumptions and constraints were proposed for optimization of bike sharing scheduling. On this basis, a decision model for bike sharing in peak hours was established with fixed time window, with the aim to minimize the scheduling cost. Then, the genetic algorithm (GA) was improved to solve the model. Our model and the GA were proved feasible through a case analysis. This research provides a reference for optimizing the operation and management of bike sharing, and promotes the public transport in large cities.