Early Detection of Emotional Behavior in Digital Era

The widespread integration of digital technology has led to a rising yet often unnoticed public health concern: internet addiction and excessive screen scrolling. These behaviors adversely affect both adolescents and adults, contributing to reduced self-esteem, emotional distress, and early symptoms of depression. Because these effects develop gradually, early detec- tion is crucial. This study introduces a data-driven framework that integrates objective digital behavior metrics with validated psychometric assessment to identify addiction risk. Digital usage data were collected using the YourHour application, which captured daily screen time, app-specific usage, phone unlock frequency, late-night activity, and habit-formation patterns. To evaluate psychological factors, participants completed a struc- tured psychometric test. The validity of questionnaire items was confirmed using multiple statistical measures: McDonald’s Omega values above 0.70 ensured internal consistency; Item Response Theory discrimination parameters (a > 1.0) and Item Discrimination Index values above 0.40 identified highly effective items; and Kaiser–Meyer–Olkin values exceeding 0.80 verified sampling adequacy and construct validity. Machine learning enhanced analytical precision. Principal Component Analysis (PCA) isolated the most influential digital and psychological predictors—such as total screen time, late-night usage, unlock frequency, emotional instability, impulsivity, and control-related traits—by retaining components with eigenvalues greater than 1. These refined features were then used by a Support Vector Machine (SVM) classifier to categorize users into minimal, moderate, or high addiction-risk levels. A pilot study involving students aged 12–15 demonstrated the system’s ability to detect emotional vulnerability and provide personalized preventive rec- ommendations. Overall, this integrated approach offers a scalable model for early intervention in digital addiction.