Article
Author: Cheng Junru Email Cheng Junru.
The rapid growth of social media usage has transformed digital communication but has also led to rising concerns about digital addiction and its impact on mental health and productivity. This study aims to predict digital addiction levels and identify key behavioral and psychological risk factors using machine learning techniques. A dataset of 1,000 social media users containing demographic, behavioral, and psychological variables was analyzed to build predictive models. Two algorithms, Random Forest and Gradient Boosting, were applied to evaluate predictive performance and interpret the determinants of addiction. The Gradient Boosting model achieved the highest accuracy, with an R² value of 0.9996, indicating its strong ability to model complex behavioral patterns. Feature importance analysis revealed that self-control, satisfaction, and productivity loss were the dominant predictors of digital addiction, while demographic factors such as age, gender, and platform type had minimal influence. The findings suggest that digital addiction is primarily driven by behavioral reinforcement and deficits in self-regulation rather than by demographic characteristics. These results contribute to a deeper understanding of digital well-being by integrating behavioral science and artificial intelligence. The study provides practical implications for designing interventions and digital policies that encourage mindful engagement, promote self-regulation, and mitigate the negative effects of excessive social media use.
Keywords: Digital Addiction, Machine Learning, Self Control, Digital Weelbeing, Social Media
How to Cite: Junru, C. (2026) “A Machine Learning Approach for Predicting Digital Addiction and Exploring Risk Factors of Social Media Overuse ”, Journal of Digital Society. 2(1). doi: https://doi.org//JDS.146