PREDICTING DIGITAL ADDICTION PATTERNS WITH MACHINE LEARNING FOR PERSONALIZED MENTAL HEALTH SUPPORT

The prevalence of digital addiction in the 21st century has become a mental health issue, especially among the younger generations who have witnessed a time of unrelenting Internet connectivity and ownership of a smartphone. This paper aims to present a comprehensive analysis of the possible application of machine learning techniques for predicting digital addiction patterns and providing personalized mental health support. This research aims to provide a synthesis of the results from multiple investigations that used supervised learning algorithms such as categorical boosting, random forest, support vector machines and deep learning architectures to classify and predict problematic digital behaviors, conducting a systematic review of the literature and methodological advances of recent empirical studies published. Evidence suggests that ensemble methods such as CatBoost and random forest are able to maintain good predictive power in different populations with ROC-AUC ranging from 0.78 to 0.93. Findings from various studies have shown that key predictive factors are screen time, sleep disruption, social media use rate, anxiety, depression and stress. Moreover, this paper investigates the incorporation of reinforcement learning parameters and behavioral inhibition systems as a computational measure of addiction vulnerability: it is shown that model-based decision making deficits and increased learning rates are strong predictors of internet addiction severity (Young, 1998). Clustering-based risk profiling and digital phenotyping provide a way to explore the translation of these predictive models to personalized intervention frameworks. Ethical issues around privacy, algorithmic bias and the possibility of pathologising normal behavior are discussed in detail. This paper ends by suggesting an integrative approach, integrating real-time behavioral monitoring, longitudinal analysis of data, and personalized feedback mechanisms, providing a blueprint for future research and clinical practice in the digital mental health field.