Risk level prediction for problematic internet use: A digital health perspective

Area of Science:

• Digital Health
• Computational Psychology
• Behavioral Science

Background:

• Problematic Internet Usage (PIU) is a growing concern across disciplines.
• Existing research has explored various theoretical and empirical approaches to PIU.
• Gaps in understanding and detection necessitate advanced methodologies.

Purpose of the Study:

• To systematically review PIU literature, identifying research trends, methodologies, and challenges.
• To develop and validate a comprehensive machine learning pipeline for improved PIU detection.
• To identify key predictive features influencing PIU detection across different severity levels.

Main Methods:

• Systematic literature review of PIU research.
• Development of a machine learning pipeline including preprocessing, feature extraction, and modeling.
• Performance validation strategies addressing missing values and data imbalance.

Main Results:

• Machine learning model performance significantly improved by handling missing data and imbalance.
• Identified key predictive features: physiological indicators, physical activity, sleep quality, and Internet usage patterns.
• Elucidated the differential impact of features on PIU detection at varying severity levels.

Conclusions:

• The developed pipeline offers enhanced understanding and detection of PIU.
• Findings provide actionable insights for early detection and tailored interventions.
• Contributes to digital health by informing internet addiction prevention and intervention programs.