Analysis of a survey for a bike sharing company

This thesis explores the role of machine learning (ML) in analyzing customer satisfaction within a bike-sharing system in Bologna, Italy. A comprehensive web-based survey of over 10,000 respondents was conducted, capturing diverse socio-demographic traits and detailed feedback on service usage, efficiency, reliability, and accessibility. Using this dataset, clustering techniques were applied to identify distinct customer groups based on satisfaction levels, with K-means emerging as the most effective method for segmentation. Further, predictive models were developed to classify users into these clusters based on demographic and behavioral factors. Random Forest outperformed other models in terms of sensitivity, providing valuable insights into the factors driving satisfaction and dissatisfaction. The findings highlight the potential of ML to enhance decision-making in shared mobility systems by uncovering hidden patterns and predicting user behavior. Clustering revealed two primary groups, allowing for tailored strategies to improve service delivery. Predictive modeling demonstrated the feasibility of identifying at-risk users for targeted interventions. By integrating ML techniques with robust survey analysis, this study provides actionable recommendations for optimizing bike-sharing systems, enhancing user satisfaction, and promoting sustainable urban mobility. Future research could explore advanced ML techniques and dynamic data sources to refine these models further, ensuring their relevance in rapidly evolving urban landscapes.