Insights on exploring a small-scale electric bicycle sharing data set

Cities can benefit to a great extent from the deployment of bicycle-sharing services (BSS). Conceived as being an alternative for taxi services and public transport, BSSs are used by daily commuters, as well as tourists. Open data sharing policies in the US allow data scientists to make use of relatively detailed and anonymized data extracts of often station-based, bicycle-sharing services. Origin destination matrices (OD matrices) allow the representation of flows while leaving out detailed information on the specific trajectory and the traffic control that influences individual movements. This paper aims at reviewing the existing methods for analyzing sharing services that operate predominantly using GPS data. We aim to use unsupervised clustering learning techniques to analyze GPS data and develop insights. We found that combining more than one clustering technique is more effective as compared to individual techniques. We discuss the merits and demerits of individual and combined techniques and their relevance in analyzing bike-sharing GPS data.