Predictive Boosting for Employee Retention with SMOTE and XGBoost Hyperparameter Tuning

Companies today face challenges in understanding the factors that influence employee retention, especially in handling imbalanced datasets that depict low retention rates. This research aims to explore the impact of various variables on employee retention rates and improve predictive capabilities through a combination of SMOTE for data imbalance handling and hyperparameter tuning in the XGBoost model. The research method involved detailed analysis of employee profiles and application of SMOTE techniques to harmonize data imbalance, along with fine tuning of hyperparameters in the XGBoost model to improve model performance. The results of the first experiment using the XGBoost model showed that the model produced an accuracy rate of 0.81, but after going through the hyperparameter tuning process, the accuracy increased significantly to 0.85. The results of this study provide a deeper understanding of the factors that contribute to employee retention in different types of companies. The implementation of SMOTE and hyperparameter tuning in the XGBoost model proved its effectiveness in improving prediction accuracy. This research has the potential to provide strategic guidance for human resource management in improving employee retention in the future.