Perbandingan Kinerja Algoritma Random Forest Dan Decision Tree Dalam Klasifikasi Tingkat Kecanduan Game

Abstract

This study compares the performance of Random Forest and Decision Tree algorithms in classifying gaming addiction levels (Mild, Moderate, Severe) using the Gaming Addiction Dataset from Kaggle, consisting of 234 valid records and 44 behavioral, psychological, and demographic features after data leakage removal. Both models were evaluated using accuracy, precision, recall, and F1-score, validated through 5-fold and repeated cross-validation along with additional overfitting analysis using regularization and Out-of-Bag scoring. The results show that Random Forest consistently outperformed Decision Tree, achieving a mean cross-validation accuracy of 84.6% compared to 77.8%, and a mean macro F1-score of 0.570 compared to 0.526, with daily playtime hours, total screen time, and dopamine dependency index identified as the most influential predictors, while both models struggled to classify the minority "Severe" class due to extreme class imbalance, highlighting a key limitation and direction for future research with larger, more balanced datasets.