PLAYER BEHAVIOUR PREDICTION IN GAME PURCHASE USING ML
- Version
- Download 2
- File Size 490.33 KB
- File Count 1
- Create Date 4 June 2025
- Last Updated 4 June 2025
PLAYER BEHAVIOUR PREDICTION IN GAME PURCHASE USING ML
Authors:
B.RUPADEVI1, SHAIK DAVOOD EBRAHIM2
1Associate Professor, Dept of MCA, Annamacharya Institute of Technology & Sciences, Tirupati, AP, India, Email:rupadevi.aitt@annamacharyagroup.org
2Post Graduate, Dept of MCA, Annamacharya Institute of Technology & Sciences, Tirupati, AP, India, Email:shaikDavoodebrahim @gmail.com
Abstract: The gaming industry increasingly relies on predictive analytics to enhance player engagement and optimize in-game purchase revenue. This study develops machine learning models to predict Player Engagement Level (PEL) and Purchase Likelihood (PL) using a dataset of 5,000 player records with 33 features, encompassing gameplay, monetization, social, and demographic attributes. Through exploratory data analysis, feature selection with SelectKBest, and class balancing via SMOTE, the methodology mitigates imbalances and reduces dimensionality to eight key predictors. Four algorithms—Decision Tree, Random Forest, Logistic Regression, and XGBoost—are evaluated, with Random Forest achieving 88.63% accuracy for PEL and XGBoost attaining 99.95% for PL. A Flask-based web application, hosted locally, integrates the models with MySQL authentication, enabling interactive predictions. Despite overfitting risks and local deployment constraints, the project provides actionable insights for player retention and monetization, establishing a scalable framework for advanced gaming analytics.
Keywords—Machine Learning, Player Behavior, In-Game Purchases, Engagement Prediction, Feature Engineering, Web Deployment, Player Engagement Level (PEL), Purchase Likelihood (PL).
Download