Context-Aware Personalized Route Recommendations for Bicycle Users Using Large-Scale Mobility Data

This research addresses the challenge of personalized bike route suggestions, an essential aspect of enhancing Mobility as a Service. The focus is on developing routes that balance efficiency (speed and distance) with the user's safety, preferences, and needs, filling a gap in modern urban mobility solutions. The approach involves analyzing the Sussex-Huawei Locomotion Dataset based on individual's historical mobility patterns, using a hybrid clustering method, Density-based Medoids, to pinpoint user's most significant locations. The accuracy of these identified hotspots is validated through metrics like the Silhouette score, Davies-Bouldin index, and Calinski-Harabasz index. Subsequently, a “bikeability” network is constructed, incorporating bike-related Points of Interest and real-time road incident data. This semantically enriched graph underpins a modified Dijkstra's algorithm, tailored for context-aware route recommendations. The effectiveness of these personalized routes is evaluated against alternatives generated by Yen's K shortest paths algorithm, using criteria such as path length, travel time, incidence of bike Points of Interest, and potential safety hazards. This study contributes to the development of smarter, safer, and more personalized urban recommendations, paving the way for future advancements in sustainable urban mobility.