The final master thesis investigate behavioural patterns of the users of bike-sharing system BIXI in Montreal. A data mining approach is used by applying clustering methods on open data from BIXI, the Government of Canada and the City of Montreal. Hierarchical clustering with Ward and Gower is processed to form clusters using only the BIXI variables at first. Then, because the authors believe that contextual information has a fundamental effect on the user’s behaviour, they investigate the impact of adding such information into the clustering methodology. In the end, a multi-view clustering method is preferred for its quality of preserving the basic characteristics of simpler set of clusters (mixing less variables). Afterwards, local analysis are done over some profiles of the classes. A dynamic balancing analysis of the station using geographical representation offer good insight on the movements of the bikes associated with a cluster. Then, a complex network approach is also used to extract more topological information about the bike-sharing network made from the stations and the trips. All these transdisciplinary approaches used together produce new operational knowledge that can be useful for the operators and the logistics to overcome the redistribution problem as well as improving the quality of the service.