<![CDATA[This research was conducted to fill the gap in previous research in swimming performance prediction that rarely combines daily lifestyle aspects with training and physiological data, especially in the context of local swimming clubs in Indonesia that still rely on conventional approaches. The purpose of this research is to develop a prediction model for 50-meter freestyle swimming times using a holistic Random Forest Regressor algorithm. The method applied is a data-oriented methodology for prediction with longitudinal primary data from 10 ZNERGY Aquatic Swimming Club adolescent athletes (81 observations) collected through coach logbooks for training variables and time targets, and digital questionnaires for parents for physiological and lifestyle variables, including a specially developed Nutrition Compliance Index (IKG). The data was processed with Python (Scikit-Learn and Pandas), including pre-processing, lagged feature engineering, and model evaluation with an 80:20 split. The results show the model has high accuracy with an R² of 0.9591, MAE of 1.511 seconds, and RMSE of 2.254 seconds; The most important variable was the previous test time (45.66%), followed by training variables (33.67%), while lifestyle variables contributed little. The implication of this study is the availability of an evidence-based decision support system for coaches to design training programs objectively, optimize athlete progress, and prevent overtraining through a holistic approach that integrates multidimensional data.]]>
