<![CDATA[This study is motivated by the rapid rise of social media use among adolescents, which may adversely affect mental health by increasing the risk of stress, anxiety, and depression; thus, early detection is essential to prevent more severe outcomes. It aims to compare the performance of K-Nearest Neighbor (KNN) and Naive Bayes in detecting adolescent mental health risks and to evaluate the impact of data balancing using the Synthetic Minority Over-sampling Technique (SMOTE). A quantitative experimental design was applied, including data preprocessing, model implementation, and evaluation using 10-fold cross-validation with accuracy, precision, recall, F1-score, and AUC as performance metrics. The results show that Naive Bayes provides more stable performance with higher accuracy and precision, while KNN combined with SMOTE significantly improves recall, particularly for minority classes, indicating a trade-off between precision and recall in model selection. This study contributes a comprehensive analysis of the role of data balancing in classification performance within mental health contexts. Future work should explore ensemble and deep learning approaches and utilize larger, more diverse datasets to enhance generalizability.]]>
Copyrights © 2026
