Similarity measure for station clustering in bike sharing systems

Bike sharing systems are one of the proposed solutions for personal transportation in cities around the world. Clustering is a principal direction for research done on bike sharing systems with the purpose of reducing computational difficulty of problems associated with operating such systems. This paper proposes a new similarity measure for station clustering based on periods of low bike availability. The method described is tested on real world bike share data obtained for Citi Bike in New York. A time-based validation is done for the proposed algorithm.