A Machine Learning Framework for Digital Game Addiction Detection and Early Intervention

GAMEADDICTED: A Machine Learning Framework for Digital Game Addiction Detection and Early Intervention proposes a comprehensive and scalable approach for identifying potential digital game addiction by analyzing player behavioral patterns using machine learning techniques. The framework leverages structured gameplay data, including playtime hours, session frequency, average session duration, player progression level, and achievement metrics, to model and predict player engagement intensity.

The system is developed as a web-based application using the Flask framework, supporting secure user authentication, dataset upload, algorithm selection, and real-time prediction. Multiple supervised learning algorithms—Random Forest, XGBoost, Decision Tree, and LightGBM—are implemented and evaluated to assess their effectiveness in classifying player engagement levels. Decision Tree models provide interpretability of behavioral patterns, while ensemble-based approaches such as Random Forest and LightGBM enhance prediction robustness and accuracy. XGBoost further improves performance through gradient-boosted decision trees, enabling precise modeling of complex player behaviors.

Based on the predicted engagement category (high, medium, or low), the framework generates personalized intervention suggestions aimed at promoting responsible gaming habits, improving player retention strategies, and mitigating addiction risks. The architecture integrates MySQL for secure data management and employs joblib for efficient deployment of trained models.

Keywords

Digital Game Addiction, Player Engagement Prediction, Machine Learning, Random Forest, XGBoost, Decision Tree, LightGBM, Behavioral Analytics, Early Intervention