An Analysis of Case Representations for Marathon Race Prediction and Planning

Abstract

We use case-based reasoning to help marathoners achieve a personal best for an upcoming race, by helping them to select an achievable goal-time and a suitable pacing plan. We evaluate several case representations and, using real-world race data, highlight their performance implications. Richer representations do not always deliver better prediction performance, but certain representational configurations do offer very significant practical benefits for runners, when it comes to predicting, and planning for, challenging goal-times during an upcoming race.