Cycling Race Time Prediction

A personalized ML approach using route topology and training load.

Summary

This paper studies how to predict cycling duration for a specific rider using route topology and current fitness state, avoiding heavy dependence on hard-to-measure physics parameters.

The approach is trained on personal historical rides and evaluated as an N-of-1 study. A Lasso-based model using topology plus fitness features provides strong predictive performance, and checkpoint predictions support practical pacing decisions during a ride.

Quick Info

Back to all papers