Classifying Internet Addiction Using Machine Learning Approach: A Study Among Adolescents in Bangladesh

Area of Science:

• Adolescent Health
• Mental Health Technology
• Data Science in Healthcare

Background:

• Internet addiction (IA) poses significant risks to adolescent mental, emotional, social, and physical well-being.
• Adolescents are particularly vulnerable to online temptations due to developmental factors.
• Limited traditional research exists on IA in Bangladesh, highlighting a need for novel approaches.

Purpose of the Study:

• To identify risk factors associated with Internet addiction (IA) in adolescents.
• To leverage advanced machine learning (ML) techniques for IA classification.
• To address the research gap in understanding IA prevalence and predictors in Bangladesh.

Main Methods:

• Convenience sampling of 385 adolescents surveyed for depression (PHQ-9), loneliness (UCLA-3), and IA (IAT-20).
• Boruta feature selection identified key IA prevalence factors.
• Evaluated multiple ML classification models including SVM, DT, LR, and RF using cross-validation and ROC curves.

Main Results:

• A significant one-third (30.1%) of respondents reported IA.
• Key predictors for IA included father's education, favorite activity, loneliness, smoking status, depression, and internet usage duration.
• The SVM linear kernel model demonstrated superior performance in classifying IA, achieving high accuracy (0.819) and AUC (0.890).

Conclusions:

• Raising awareness about IA among adolescents and parents is critical due to its high prevalence.
• ML frameworks effectively identify prognostic indicators for IA, aiding in accurate classification and intervention.
• Findings can inform policy-making and counseling services to combat the adolescent IA crisis.