Prediction of factors that affect sportsman performance by using support vector machines

The aim of this study is to predict aerobic performance level of a sportsman by using Support Vector Machines (SVM), which was defined as a kind of statistical learning system. In this study; age, height, weight and test results belonging to Sportsmen that had cardiopulmonary exercise tests at Çukurova University Sport Physiology Laboratory between the years of 2003 and 2006 were used. According to the exercise test protocol, velocity and slope were increased gradually and the quantities of minute ventilation volume (VE), oxygen consumption (VO2), carbon dioxide generation (VCO2), and heart rate were saved in a certain time interval. To analyze data with SVM, software was developed by using MATLAB programming language that makes high level technical computing. The sportsmen have been separated in three groups, which were named as train group, test group and predict group respectively. 17 of them were included in the train group, 10 of them were included in the test group and 5 of them were included in the predict group. It has been shown that exercise test data become more stable by taking average of data and applying a curve-fitting algorithm to data. The decision function that was obtained after learning phase has been applied to the datasets that belong to sportsmen who were included in the test and predict group in turn in order. As a result, performance levels of all sportsmen who were included in the test group have been predicted correctly. SVM has successfully been used in a wide range of areas in analysis of many problems. Also in this study, performance levels of sportsmen were predicted with SVM successfully. Having an idea that this method can be applied in many subjects on physical education and sports area, further studies are needed that use SVM.