Bike sharing is a growing phenomenon which aims at reducing traffic congestion and gas emissions by providing a healthier means of public transportation in urban areas across the globe. Users are more likely to adopt this option if it comes with guarantees on its quality of service — one would like to find a bike at the station where the journey starts, and an empty parking spot at the end of the ride.
Researchers from the EU QUANTICOL project have developed a new technique to accurately predict the availability of bike-sharing networks using a mathematical model with “explanatory” features: It postulated specific rules of evolution that were then successfully validated on a large dataset of the Vélib’ system, covering the whole metropolitan area of the city of Paris.
This technique holds promise for improving the user experience and system efficiency. From the users’ viewpoint it can be employed as a recommendation system to rank stations according to their likelihood of being able to satisfy a request. The network operator, on the other hand, can exploit its predictive capabilities to proactively anticipate undesired situations such as the depletion of available bikes (or full occupancy of parking slots) at popular stations.
This research was recently presented at the 24th ACM International Conference on Information and Knowledge Management in Melbourne, Australia, a top-tier computer science venue promoting work at the intersection between information retrieval, knowledge management, and databases.