Investigating shared bike usage patterns in correlation with weather conditions through the application of machine learning algorithms

This Master's Thesis presents an analysis of shared bike usage based on weather conditions, utilizing machine learning algorithms capable of predicting future shared bike usage based on weather forecasts. As these services become increasingly integral to urban mobility, the reliability of supporting activities, such as accurate weather forecasts, is crucial to enhancing the overall efficiency of the system. The machine learning algorithms employed in this Thesis belong to the category of supervised learning techniques. These algorithms learn to predict the desired parameter by analyzing a vast dataset containing numerous historical examples. The dataset, in this case, spans one year of records detailing the number of bikes rented in Vicenza, along with the corresponding weather conditions during that period. Among the various algorithms explored, the random forest algorithm emerged as the most effective in providing accurate results. This study identifies an opportunity for the municipality to formulate targeted strategies promoting year-round bike usage based on weather-related patterns. The positive correlation between mean temperature, solar radiation, and extended trip durations in summer suggests a propensity for heightened bike activity during warmer and sunnier conditions. In light of these findings, initiatives such as promoting bike-sharing programs, improving bike-friendly infrastructure, and organizing events during the summer months are recommended to capitalize on this observed trend, fostering increased community engagement and sustainable transportation habits.