A Support Vector Machine Approach for Effective Bicycle Sharing in Urban Zones

Abstract

Bicycle-sharing system is a modern personalized public transport network through which people can rent a bicycle from one bicycle stand to the other in their network. It is an environment-friendly mode of transport for a healthier society. Today, the bicycle-sharing system is getting popular all over the world but inconsistent demand in different time slot and unreasonable bicycle allocation at different bicycle stand is a potential concern for any governing company. Thus, bicycle demand on a particular day and particular time slot is very important for the distribution of limited available resources. In this paper, we have predicted the bicycle demand based on a few important attributes like temperature, humidity, and time that affect the performance. We have used a support vector machine for this regression task where predictions are made both on a daily and hourly basis. So that the business administrator can effectively redistribute the bicycle across the location based on the demand, and end user can effectively plan their trip to nearby catchment areas. The evaluation result indicates that our method achieved an acceptable accuracy on both daily and hourly bicycle-sharing dataset. However, effective re-balancing at particular bicycle stand is possible if a potential destination or raw sensor data of a particular bicycle stand is known beforehand.