<![CDATA[The rapid growth in the number of online game players has generated large volumes of behavioral data that can be leveraged to analyze player engagement levels. However, the primary challenges include class imbalance in engagement data and the limited interpretability of predictive models regarding the factors influencing their decisions. This study aims to develop a classification model for online game player engagement levels using the Random Forest algorithm, while addressing class imbalance through the Synthetic Minority Over-sampling Technique (SMOTE) and identifying the most influential features using Feature Importance analysis. The study utilized the Online Gaming Behavior Dataset from Kaggle, comprising 40,034 records. The Random Forest model was optimized using RandomizedSearchCV with five-fold cross-validation to determine the optimal hyperparameter configuration. The experimental results demonstrate that the proposed model achieved an overall accuracy of 92%, with recall values of 90%, 94%, and 89% for the Low, Medium, and High engagement classes, respectively. Feature Importance analysis using both impurity-based and permutation approaches consistently identified SessionsPerWeek (0.4375 and 0.4309) and AvgSessionDurationMinutes (0.3371 and 0.3540) as the two most influential features, jointly accounting for more than 77% of the model's predictive decisions, whereas demographic features contributed only marginally. The novelty of this study lies in the integration of Random Forest, SMOTE, and Feature Importance to develop a classification model that is not only highly accurate but also interpretable. These findings provide valuable insights for game developers in designing evidence-based player retention strategies, such as implementing daily login rewards to increase session frequency and time-limited events to encourage longer gameplay sessions.]]>
