Problematic Internet Use Linked to Impulsive and Compulsive Behaviors Using Machine Learning in Psychiatry

Of the 2006 participants, 181 (9.0%) exhibited moderate to severe problematic internet use.

• Problematic internet use was identified through the Internet Addiction Test (IAT).
• Logistic Regression and Naïve Bayes models achieved a prediction accuracy with a receiver operating characteristic area under the curve (ROC-AUC) of 0.83.
• The Random Forests algorithm yielded a ROC-AUC of 0.84, indicating a strong predictive capability.
• All predictive models outperformed baseline models with statistical significance (p < 0.0001).
• The models demonstrated reliable transferability between the study sites in validation tests (p < 0.0001).
• Specific measures of impulsivity and compulsivity were effective in predicting problematic internet use.

Simplified