The purpose of the present study was to determine a model of anthropometric, physical fitness, and sport-specific performance data in young canoe sprint athletes for the prediction of adult performance level. In a prospective cohort study design, anthropometric (e.g., body height/mass), physical fitness (e.g., muscular endurance, linear speed), and sport-specific performance data (e.g., 250/2000 m on-water trials) of 731 young canoe sprint athletes were obtained between 1992 and 2019. Prediction models comprised random forest classifier, logistic regression, and naive bayes. The comparison of different prediction models showed that random forest classifier outperformed the others. More precisely, the binary outcome of adult performance level (international vs. national) was predicted with an overall accuracy of 95%. The estimated chance of belonging to either the performance level national or international can help coaches and practitioners of canoe sprint athletes for decision-making during the course of long-term athlete development and talent identification.
© Copyright 2022 Proceedings of the 9th International Performance Analysis Workshop and Conference & 5th IACSS Conference. PACSS 2021. Advances in Intelligent Systems and Computing, vol 1426. Published by Springer. All rights reserved.