Analyzing Patterns of a Bicycle Sharing System for Generating Rental Flow Predictive Models

Urban mobility has been highlighted as one of the most relevant themes in Smart Cities. Alongside this, following a principle of resource optimization and seeking greater sustainability, bicycle sharing systems have stood out as a resource that can be used to assess urban mobility. The correct analysis of these data and the understanding of the dynamics in these systems can aid in decision making, in addition to optimize the complex urban mobility system. Thus, this work analyzes a Bicycle-Sharing System dataset, which is enriched for us with meteorological and seasonal information. In order to achieve our results, we recognize cyclist activity patterns related to date and climate information, as well as we identify a set of parameters that influences bicycle rental flow. Finally, we explore the relationship between these parameters and patterns, in order to present predictive regression models for rental flow prediction. In our results, Random Forest algorithm was the best approach for the creation of an effective regression model, explaining 95% of the explanatory variables.