Support vector machines for aerobic fitness prediction of athletes


AÇIKKAR M. , AKAY M. F. , ÖZGÜNEN K. T. , AYDIN K. , KURDAK S. S.

EXPERT SYSTEMS WITH APPLICATIONS, cilt.36, ss.3596-3602, 2009 (SCI İndekslerine Giren Dergi) identifier identifier

  • Cilt numarası: 36 Konu: 2
  • Basım Tarihi: 2009
  • Doi Numarası: 10.1016/j.eswa.2008.02.002
  • Dergi Adı: EXPERT SYSTEMS WITH APPLICATIONS
  • Sayfa Sayısı: ss.3596-3602

Özet

Support vector machine is a statistical learning classifier, based on the principle of structural risk minimization, which performs well when applied to data outside the training set. This paper presents a new approach based on support vector machines to predict whether an athlete is aerobically fit or not. The input data set contains physical properties of athletes as well as their cardiopulmonary exercise testing results which were obtained at Cukurova University Sports Physiology Laboratory. According to the exercise test protocol, speed and grade of the treadmill were increased at certain times and the input variables of time, speed and grade of the treadmill, and oxygen uptake, carbon dioxide output, minute ventilation and heart rate of athletes were recorded. The average of the exercise test data was taken over certain time intervals and a curve fitting algorithm was applied to remove the spikes in the data and make it more suitable to use with support vector machines. Experiments with several different training and test data show that curve-fitted data has better performance measures, such as higher prediction rate, sensitivity, specificity, and shorter training time. (C) 2008 Elsevier Ltd. All rights reserved.

Support vector machine is a statistical learning classifier, based on the principle of structural risk minimization, which performs well when applied to data outside the training set. This paper presents a new approach based on support vector machines to predict whether an athlete is aerobically fit or not. The input data set contains physical properties of athletes as well as their cardiopulmonary exercise testing results which were obtained at Cukurova University Sports Physiology Laboratory. According to the exercise test protocol, speed and grade of the treadmill were increased at certain times and the input variables of time, speed and grade of the treadmill, and oxygen uptake, carbon dioxide output, minute ventilation and heart rate of athletes were recorded. The average of the exercise test data was taken over certain time intervals and a curve fitting algorithm was applied to remove the spikes in the data and make it more suitable to use with support vector machines. Experiments with several different training and test data show that curve-fitted data has better performance measures, such as higher prediction rate, sensitivity, specificity, and shorter training time.