KMAN Publication Inc. (KMANPUB) Journal of Adolescent and Youth Psychological Studies (JAYPS) 2981-2526 7 Serial Number 46 2026 06 10 K-Nearest Neighbors Classification of High-Risk Internet Addiction Profiles among Adolescents Based on Impulsivity, Loneliness, Sleep Quality, and Parental Monitoring K-Nearest Neighbors Classification of High-Risk Internet Addiction Profiles among Adolescents Based on Impulsivity, Loneliness, Sleep Quality, and Parental Monitoring 1 13 10.61838/kman.jayps.5432 EN Journal Article 2026 02 05 <table> <tbody> <tr> <td> <p><strong>Objective:</strong> This study aimed to develop and evaluate a K-Nearest Neighbors classification model for identifying high-risk internet addiction profiles among Canadian adolescents based on impulsivity, loneliness, sleep quality, and parental monitoring.</p> <p><strong>Methods and Materials:</strong> This cross-sectional predictive study was conducted among 1,284 adolescents aged 13 to 18 years from secondary schools in Canada. Participants completed standardized self-report measures, including the Internet Addiction Test, Barratt Impulsiveness Scale–11, UCLA Loneliness Scale Version 3, Pittsburgh Sleep Quality Index, and Parental Monitoring Scale. Internet addiction risk status was defined using established Internet Addiction Test cutoff scores, and adolescents were classified into high-risk and low-risk groups. Data preprocessing included missing-value treatment, outlier screening, and standardization of continuous predictors. The dataset was divided into training and testing subsets using stratified sampling. A K-Nearest Neighbors algorithm was trained to classify internet addiction risk, and hyperparameter optimization was performed using grid search with five-fold cross-validation.</p> <p><strong>Findings:</strong> The optimized KNN model demonstrated strong inferential and predictive performance in classifying adolescent internet addiction risk. The final model achieved an accuracy of 88.72%, precision of 85.91%, recall of 83.47%, specificity of 90.84%, F1-score of 84.67%, AUC-ROC of 0.924, and Cohen’s kappa of 0.742. The confusion matrix showed that the model correctly classified 171 low-risk adolescents and 57 high-risk adolescents, with 17 false positives and 12 false negatives. Hyperparameter optimization indicated that the strongest model performance was obtained with k = 11 using the Euclidean distance metric. Permutation importance analysis identified impulsivity as the strongest predictor, followed by parental monitoring, loneliness, and sleep quality.</p> <p><strong>Conclusion:</strong> The findings indicate that K-Nearest Neighbors classification can accurately identify high-risk internet addiction profiles among adolescents using a concise set of psychological, sleep-related, and family-based predictors.</p> </td> </tr> </tbody> </table> https://journals.kmanpub.com/index.php/jayps/article/download/5432/10535