Penerapan Model Regresi Logistik Biner dan Random Forest terhadap Prospek Atlet Muda pada Liga Basket DBL Tahun 2019

Abstract

Athletic performance is a key indicator of the success of athlete development within a sports discipline, including basketball. Effective development requires a competitive and professional platform for talent identification, such as the DBL East Java Series basketball league for senior high school students. A well-organized competition supports positive athlete development, enabling the evaluation of individual prospects through game statistics. Athlete prospects reflect future potential arising from present performance and are categorized according to the Indonesia Emas (PRIMA) Program Guidelines established by the Indonesian National Sports Committee in 2015, which define the Pratama class for athletes competing at national or regional levels. This study develops classification models to predict athlete prospects using match-level statistics. To address class imbalance, the Synthetic Minority Oversampling Technique (SMOTE) is applied, while k-fold cross-validation is used to obtain robust model estimates. The findings show that all constructed models achieve strong predictive performance based on the Area Under the Curve (AUC). Furthermore, the variables points scored—representing scoring ability—and assists—representing ball-handling and playmaking ability—are identified as the most influential predictors of young athlete prospects in both the binary logistic regression and random forest models.