No abstract

Maximum oxygen consumption (VO2max) is important to observe the endurance of the athletes and evaluate their performance.. Aim is to develop new prediction models for college-aged students using Support Vector Machine (SVM) with Relief-F feature selection algorithm. Ten different models consisting of the predictor variables gender, age, weight, height, maximal heart rate (HRmax), time, speed, Perceived Functional Ability scores (PFA-1 and PFA-2) and Physical Activity Rating score (PA-R) have been created by Relief-F scores for prediction of VO2max. The prediction models’ standard error of estimates (SEE’s) and multiple correlation coefficients (R’s) have been calculated for evaluating their performances. For comparison purposes, Tree Boost (TB) and Radial Basis Function Network (RBFN) based models have also been developed. The results show that the prediction model including PAR, speed, time, weight, PFA-1, gender and HRmax gives the lowest SEE with 6.42 mL.kg−1.min−1 and highest R with 0.79. Also, this study shows that the predictor variables HRmax and gender play a considerable role in VO2max prediction. Keywords: Maximum oxygen uptake, machine learning, feature selection.

No abstract

Quadriceps refers to a group of four muscles on the front of the thigh. Adequate quadriceps strength is essential for athletic performance. Quadriceps strength in athletes can be reliably assessed using isokinetic dynamometry. Also, some studies in literature showed the possibility of predicting the quadriceps strength of athletes using machine learning methods within acceptable error rates. The purpose of this study is to investigate the effect of sport branch on quadriceps strength prediction using Support Vector Machine (SVM). The dataset included 70 athletes selected from the College of Physical Education and Sport at Gazi University. The optimal values of SVM parameters have been found by using grid search. The predictor variables gender, age, height, weight and sport branch have been utilized to build sixteen different quadriceps strength prediction models. By carrying out 10-fold cross-validation, the performance of the prediction models has been evaluated by calculating the root mean square errors (RMSE's) and multiple correlation coefficients (R's). The results show that the RMSE's of the prediction models change from 23.31 to 47.78 Nm. The model including the predictor variables gender, height, weight and sport branch yields the lowest RMSE and highest R. One can conclude that sport branch has a profound effect for predicting the quadriceps strength of athletes.

Özetçe-Üst vücut gücü (Upper BodyPower -UBP), kros kayakçıların yarış performansını belirleyen en önemli unsurlardan biridir. Bu çalışmada farklı makine öğrenme yöntemleri ile kros kayakçılara ait 10 saniye UBP (UBP 10 ) ve 60 saniye UBP (UBP 60 ) değerlerini tahmin etmek üzere yeni modeller geliştirilmiştir. Makine öğrenme yöntemleri olarak Kademeli Korelasyon Ağı (Cascade Correlation Network -CCN), Radyal Tabanlı Fonksiyon Sinir Ağı (Radial Basis Function Neural Network -RBF) ve Karar Ağacı Ormanı (Decision Tree Forest -DTF) kullanılmıştır. Modellerde tahmin değişkenleri olarak cinsiyet, yaş, vücut kitle indeksi (body mass index -BMI), kalp atış hızı (heart rate -HR), maksimum oksijen tüketimi (VO 2 max) ve egzersiz süresi kullanılmıştır. Tahmin modellerinin performansı, 10 katlı çapraz doğrulama kullanılarak, çoklu korelasyon katsayısı (Multiple Correlation Coefficient -R) ve standart tahmin hatası (Standard Error of Estimate -SEE) hesaplanarak değerlendirilmiştir. Elde edilen sonuçlar incelendiğinde; yaş, cinsiyet, BMI ve VO 2 max değişkenlerini içeren CCN tabanlı UBP 10 ve UBP 60 tahmin modellerinin en düşük SEE değerlerini ürettiği gözlemlenmiştir. Anahtar Kelimeler -makine öğrenimi; üst vücut gücü; regresyon. Abstract-Upper Body Power (UBP) is one of the most important determinants that directly affects the performance of cross-country skiers during races. In this study, new models have been developed to predict the 10-second UBP (UBP 10 ) and 60second UBP (UBP 60 ) of cross-country skiers by using different machine learning methods including Cascade Correlation Network (CCN), Radial Basis Function Neural Network (RBF)and Decision Tree Forest (DTF). The predictor variables used to develop prediction models are age, gender, body mass index (BMI), heart rate (HR), maximal oxygen uptake (VO 2 max) and exercise time. By using 10-fold cross-validation on the dataset, the performance of the prediction models has been evaluated by calculating their multiple correlation coefficient (R) and standard error of estimate (SEE). The results show that the CCN-based model including the predictor variables age, gender, BMI and VO 2 max yields the lowest SEE both for the prediction of UBP 10 and UBP 60 .