Predicting the maximum endurance time for left-side bridge exercise using machine learning methods and hybrid data

Abstract

This study was carried out with the intention to create new models to predict the maximum endurance time for the left-side bridge exercise using machine learning methods and hybrid data. Particularly, four different methods including Multilayer Feed-Forward Artificial Neural Network (MFANN), Generalized Regression Neural Network (GRNN), Radial Basis Function Neural Network (RBFNN) and Single Decision Tree (SDT) have been used for model development. The dataset used to create the prediction models includes physiological, exercise and questionnaire data related to individuals who performed the left-side bridge exercise and completed the Perceived Activity Rating (PAR) and Perceived Functional Ability (PFA) questionnaires. To evaluate the performance of the models, two well-known metrics, namely Root Mean Square Error (RMSE) and Multiple Correlation Coefficient (R) have been used, whereas the generalization errors have been assessed using 10-fold cross validation. The best prediction performance among the models has been obtained by using MFANN along with the predictor variables gender, age, body mass index (BMI), the times to reach a rate of perceived exertion values of 7 and 8 (RPE-7 and RPE-8, respectively) and PAR, producing the lowest RMSE and the highest R with 10.61 seconds (s) and 0.92, respectively.