<abstract xmlns="http://www.w3.org/1999/xhtml"><p> Two different computational approaches were used to predict Olympic distance triathlon race time of German male elite triathletes. Anthropometric measurements and two treadmill running tests to collect physiological variables were repeatedly conducted on eleven male elite triathletes between 2008 and 2012. After race time normalization, exploratory factor analysis (EFA), as a mathematical preselection method, followed by multiple linear regression (MLR) and dominance paired comparison (DPC), as a preselection method considering professional expertise, followed by nonlinear artificial neural network (ANN) were conducted to predict overall race time. Both computational approaches yielded two prediction models. MLR provided R² = 0.41 in case of anthropometric variables (predictive: pelvis width and shoulder width) and R² = 0.67 in case of physiological variables (predictive: maximum respiratory rate, running pace at 3-mmol·L<sup>-1</sup> blood lactate and maximum blood lactate). ANNs using the five most important variables after DPC yielded R² = 0.43 in case of anthropometric variables and R² = 0.86 in case of physiological variables. The advantage of ANNs over MLRs was the possibility to take non-linear relationships into account. Overall, race time of male elite triathletes could be well predicted without interfering with individual training programs and season calendars.</p></abstract>

Two different computational approaches were used to predict Olympic distance triathlon race time of German male elite triathletes. Anthropometric measurements and two treadmill running tests to collect physiological variables were repeatedly conducted on eleven male elite triathletes between 2008 and 2012. After race time normalization, exploratory factor analysis (EFA), as a mathematical preselection method, followed by multiple linear regression (MLR) and dominance paired comparison (DPC), as a preselection method considering professional expertise, followed by nonlinear artificial neural network (ANN) were conducted to predict overall race time. Both computational approaches yielded two prediction models. MLR provided R² = 0.41 in case of anthropometric variables (predictive: pelvis width and shoulder width) and R² = 0.67 in case of physiological variables (predictive: maximum respiratory rate, running pace at 3-mmol·L-1 blood lactate and maximum blood lactate). ANNs using the five most important variables after DPC yielded R² = 0.43 in case of anthropometric variables and R² = 0.86 in case of physiological variables. The advantage of ANNs over MLRs was the possibility to take non-linear relationships into account. Overall, race time of male elite triathletes could be well predicted without interfering with individual training programs and season calendars.

Predicting Elite Triathlon Performance: A Comparison of Multiple Regressions and Artificial Neural Networks

International Journal of Computer Science in Sport

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.

Two different computational approaches were used to predict Olympic distance triathlon race time of German male elite triathletes. Anthropometric measurements...

Predicting Elite Triathlon Performance: A Comparison of Multiple Regressions and Artificial Neural Networks

Article Category: Special Issue on Modeling in Endurance Sports Movement and Sport Science, Editors: Chris R. Abbiss, Dietmar Saupe

Published Online: Nov 30, 2017

Page range: 101 - 116

DOI: https://doi.org/10.1515/ijcss-2017-0009

Keywords
professional triathletes, performance prediction, racetime normalization, factor analysis, dominance paired comparison

© 2017

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.

Predicting Elite Triathlon Performance: A Comparison of Multiple Regressions and Artificial Neural Networks

Article Category: Special Issue on Modeling in Endurance Sports Movement and Sport Science, Editors: Chris R. Abbiss, Dietmar Saupe

Published Online: Nov 30, 2017

Page range: 101 - 116

DOI: https://doi.org/10.1515/ijcss-2017-0009

Keywordsprofessional triathletes, performance prediction, racetime normalization, factor analysis, dominance paired comparison

© 2017

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.

Keywords
professional triathletes, performance prediction, racetime normalization, factor analysis, dominance paired comparison