Drift Detection for Identifying Training Patterns Prior to Performance Improvement in Runners

The growing popularity of running in Brazil has led to an increase in amateur runner participation, creating a demand for personalized recommendations to support these athletes. This study aimed to identify runners who showed performance improvements by detecting concept drift in their training performance time series. Training metrics were computed and compared between periods preceding performance improvements and declines. It was observed that, in cases of improvement, these metrics were higher compared to the periods preceding performance declines. These differences were statistically significant (p-value < 0.05), indicating which training patterns are associated with performance progression and representing a first step toward deeper insights.