[4Xin2] Poster session 2

[4Xin2-58] Prediction of a Significant Slowing of Pace in the Latter Half of a Full Marathon

based on kinetic and kinematic characteristics in the first half

〇Yosuke Miyazaki¹, Satoru Abe¹ (1.ASICS Corporation)

Keywords:Running, Imbalanced Data, Pace analysis, IMU

“Hitting the wall” is a well-known phenomenon among runners, characterized by an unexpected slowdown in pace during the latter half of a full marathon. While several studies have described the characteristics of runners who encounter this, few have attempted to predict its occurrence. This study aimed to build a prediction model using running form characteristics, including their changes and variability, during the early stages of the race. These characteristics were obtained using a wearable inertial measurement unit. The dataset included 5,953 thirty-kilometer races for model building and 336 full marathon races for evaluation. Logistic Regression, Random Forest, and Neural Network models all surpassed random predictions in terms of PR-AUC, precision, and positive likelihood ratio, while the recall of these models remained at a level comparable to random predictions. These results suggest that running form characteristics may serve as partial indicators of pace collapse.