Machine Learning in HR Analytics: A Comparative Study on the Predictive Accuracy of Attrition Models

One of the biggest challenges in human resource management is forecasting staff loss. This study will compare ML models. The research identifies the best accurate forecasting techniques to help firms retain key personnel and improve staff stability and performance. The statistical analysis used a comprehensive dataset including demographic information, work characteristics, and satisfaction indices, all of which have the potential to influence employee attrition. We used several machine learning models, such as Logistic Regression, Decision Tree, Random Forest, GBM, XGBoost, SVM, and KNN, to forecast attrition. With an emphasis on the underrepresented group of departing employees, we compared each model's accuracy, precision, recall, and F1-score. The area of human resource analytics is substantially improved by this study's thorough analysis of several machine learning algorithms for attrition prediction. The model comparison and performance evaluation conducted in this study provide valuable insights for practitioners and scholars. The article enhances the subject by examining overlooked ensemble and boosting approaches in HR analytics literature and comparing them to mainstream models. These sophisticated models are capable of capturing complex employee attrition patterns that simpler models may fail to detect. Model performance exhibits variability, with ensemble methodologies such as Random Forest and Gradient Boosting Machines demonstrating superior predictive capabilities for staff retention. When focusing on the accuracy and recall of the minority class (workers who left), the SVM model demonstrated a significant equilibrium, highlighting its usefulness in predicting attrition.