<![CDATA[The digital era has made smartphones an inseparable part of students' lives, but it also raises the risk of addiction that negatively impacts academic achievement and mental health. This research aims to develop and evaluate machine learning models capable of classifying the level of smartphone addiction among Esa Unggul University students. Data were collected from 32 student respondents through an online questionnaire adopting the validated psychometric instrument, the Smartphone Addiction Scale-Short Version (SAS-SV). Addiction levels were categorized into two classes: 'High', which refers to the gender-specific addiction risk threshold from Kwon et al. (2013), and 'Moderate', which includes scores below that threshold. Four classification algorithms—Logistic Regression, K-Nearest Neighbors (KNN), Decision Tree, and Random Forest—were implemented to compare their performance. To address class imbalance in the data, the SMOTE oversampling technique was applied to the training data. Model evaluation was based on accuracy, precision, recall, and F1-score. The research results show that the Logistic Regression model achieved the best performance with an accuracy of 1.0000 and an average F1-score of 1.00 on the test data. Nevertheless, it should be noted that this perfect performance was obtained from a very limited test data size (8 samples), so generalization requires further validation. Feature importance analysis from the Random Forest model identified that the question related to Planned tasks/work often interrupted by smartphone use (Q0) was the most dominant predictor. These results indicate that machine learning models based on psychometric scales have initial potential as a screening and exploratory tool to identify students at risk of smartphone addiction, but require extensive development and validation on larger data before practical implementation.]]>
