Application of machine learning in predicting adolescent Internet behavioral addiction

Area of Science:

• Adolescent psychology
• Behavioral science
• Machine learning applications

Background:

• Internet addiction is a growing concern among adolescents.
• Identifying risk factors is crucial for developing effective prevention strategies.

Purpose of the Study:

• To explore risk factors associated with adolescent Internet addiction.
• To develop and evaluate machine learning models for predicting Internet addiction.

Main Methods:

• Stratified cluster sampling of 4461 high school students in Chongqing.
• Logistic regression analysis to identify independent risk factors.
• Comparison of six machine learning models (MLP, RF, KNN, SVM, LR, XGBoost) for prediction.

Main Results:

• Male gender, urban residence, and specific psychological traits (e.g., psychoticism, neuroticism, depression) were significant risk factors.
• Extreme gradient boosting (XGBoost) demonstrated the highest predictive performance (AUC=0.836) among the evaluated models.
• Machine learning models showed moderate overall performance in predicting adolescent Internet addiction.

Conclusions:

• Adolescent Internet addiction is linked to male gender, urban living, and certain personality traits.
• Machine learning, especially XGBoost, offers a promising approach for predicting and managing adolescent Internet addiction.
• Findings can inform targeted interventions for at-risk adolescents.