Jurnal Komtika (Komputasi dan Informatika)

The increased use of social media among adolescents has been identified as a suspected contributing factor to the emergence of mental health issues, including depression. However, identifying depression risk in adolescents remains challenging, primarily due to data imbalanced, which can affect the performance of classification models. This study aims to build a classification model for adolescent depression risk based on social media usage frequency using the naive bayes algorithm, and compare classification performance before and after applying SMOTE based on accuracy, precision, recall, and F1-Score. This study utilized the Teen Mental Health Dataset obtained from Kaggle. The research stages include data pre-processing, feature and target separation, splitting the data into training and testing sets with an 80:20 ratio. Applying the SMOTE method to the training data, and classification using the naïve bayes algorithm. Model evaluation was conducted using a confusion matrix. The test result showed that the model without SMOTE achieved an accuracy of 99%, precision of 100%, recall of 50%, and an F1-Score 67%. Meanwhile, the model with the application of SMOTE achieved an accuracy 97%, precision of 50%, recall of 33%, and an F1-Score of 40%. Based on the research results, the naïve bayes algorithm is capable of providing excellent classification performance in prediction the risk of depression in adolescents. The application of SMOTE successfully balanced the class distribution in the training data, but did not provide a significant performance improvement compared to the model without SMOTE. Therefore, the naïve bayes model without SMOTE is the best approach for the dataset used in this study.