The construction of models aiming the identification of the best predictors of swimming performance at each age level is one of the most pursued issues of swimming research. In this study, different approaches have been used, including Data Mining algorithms and multi-regression analysis. Anthropometric, strength, flexibility, metabolic, hydrodynamic and experience of training data of 494 young national level swimmers (13 to 16 years old boys and 12 to 14 years old girls) were measured. Best predictors were mean velocity in "T30" aerobic test for all age groups, plus height and maximal isometric elbow extension strength for older female swimmers, and glide after start for younger male swimmers. The explanation of performance reached almost 70%. The most predictable variables were similar when using different algorithms like Decision Tree, K-Means (5-clusters) and Kohonen.
© Copyright 2006 Biomechanics and Medicine in Swimming X. Veröffentlicht von Faculdade de Desporto da Universidade do Porto. Alle Rechte vorbehalten.