Predicting Player Engagement Levels in Online Gaming Platforms Using Machine Learning and Behavioral Interaction Data Analysis

Online gaming platforms have become an important component of digital society, generating large volumes of behavioral data that can be analyzed to understand player engagement patterns. This study aims to predict player engagement levels in online gaming environments using machine learning techniques. The dataset used in this research consists of 40,034 player records with 13 variables that include demographic characteristics and gameplay behavior indicators such as playtime hours, number of weekly sessions, average session duration, player level, and achievements unlocked. The target variable classifies engagement into three categories: Low, Medium, and High engagement. Several machine learning models were evaluated, including Gradient Boosting, Neural Network, and XGBoost, and their performance was assessed using accuracy, precision, recall, and F1 score. The experimental results show that XGBoost achieved the best predictive performance with an accuracy of 91.4 percent and an F1 score of 0.9137. Cross validation results further confirm the robustness of the model with a mean F1 score of 0.9163. Feature importance analysis reveals that SessionsPerWeek and AvgSessionDurationMinutes are the most influential factors in predicting player engagement, followed by achievements unlocked and player level. These findings indicate that engagement in online gaming platforms is primarily influenced by behavioral interaction patterns rather than demographic characteristics. The results highlight the potential of machine learning methods for analyzing human engagement behavior in digital environments and provide insights that may support the development of more engaging online gaming platforms.