Data-driven smart bike-sharing system by implementing machine learning algorithms

This paper aims to solve a real-life problem: the bike-sharing management system arises the requirement of offering the customers the accessibility of the bikes in different bike-stations concerning the potential demands in every time-slice. The prediction of needs is critical to the distribution of the limited resources (bikes and empty slots to place the bikes) and the management of the system. We propose addressing this problem by using the regression model, which is trained by the raw data collecting from the different sensors. Thanks to the wide distribution of the edge devices, the machine learning algorithms, and the advanced computing ability, we may incorporate the intelligence to the database-related system. We will demonstrate that the boosting gradient method as a predictor to forecast the quantities of rentals and returns of bikes, outperforming the other means, e.g., random forest, support vector machine, etc. It reaches a promising result; the average accuracy reaches 75%.