Factors influencing docked bike-sharing usage in the City of Kigali, Rwanda

Abstract

Over the past years, bike-sharing programs have evolved and passed through various developmental stages since 1965, becoming a significant part of urban mobility worldwide. Researchers conducted numerous studies to examine the usage of bike-sharing systems. While earlier research has highlighted the benefits of bike-sharing, limited attention has been given to changes in docked bike-share systems and the use of machine learning algorithms to predict docked bike-sharing usage. This research investigated the effectiveness of machine learning models in predicting docked bike-sharing station usage in Kigali City. Descriptive statistics are analysed to reveal user characteristics by Gender, education, age, and occupation. The Random Forest Model effectively classified docked bike-sharing users and non-users, achieving a balanced accuracy of 84 %. With a sensitivity of 75 % and an F1 score of 82.5 %, it demonstrated strong user identification while balancing precision and recall and a positive predictive value of 91.6 %. The study also examined the factors influencing program usage. Results indicated that Gender positively affects docked bike-sharing, with a slightly higher impact from male users. Specific stations are popular among students, while others attract non-students. Corridor analysis revealed that the Central Business District positively impacts docked bike-sharing usage. Temporal and spatial trends indicate higher usage during school months, with younger riders dominating the age distribution of users. Demand also varies by season. This study provides valuable insights to support the optimisation of docked bike-sharing operations and to guide city planners in developing relevant infrastructure and policies.