Time-aware analysis of overlapping cycling segments from activity tracker data

Artificial Intelligence is used increasingly in sports to extract actionable insights and enhance athlete performance. The widespread adoption of sports activity trackers has resulted in large volumes of detailed training data. This paper presents an improved method for analyzing cycling data from TCX files. Overlapping path segments are detected across sessions using spatial filtering and Fréchet distance, and data fusion with OpenStreetMap provides contextual information. An experiment explores how a cyclist’s performance varies on shared segments, depending on the time of day, revealing temporal patterns in training behavior.