Machine learning tools in the analyze of a bike sharing system

Abstract

Advanced models, based on artificial intelligence and machine learning, are used here to analyze a bike-sharing system. The specific target was to predict the number of rented bikes in the Nova Mesto (Slovenia) public bike share scheme. For this purpose, the topological properties of the transport network were determined and related to the weather conditions. Pajek software was used and the system behavior during a 30-week period was investigated. Open questions were, for instance: how many bikes are shared in different weather conditions? How the network topology impacts the bike sharing system? By providing a reasonable answer to these and similar questions, several accurate ways of modeling the bike sharing system which account for both topological properties and weather conditions, were developed and used for its optimization.