Comparative Analysis of Employee Attrition Prediction System Models using Machine Learning and SMOTE

Abstract

Employee attrition poses significant challenges to organizations, affecting productivity, operational costs, and workforce stability. Predicting attrition and employee behavior through machine learning (ML) models offers a data-driven approach to developing effective retention strategies. This study compares the performance of multiple ML algorithms, including Logistic Regression, Random Forest, and others, both before and after applying the SMOTE (Synthetic Minority Over-sampling Technique) boosting algorithm to address class imbalance. The analysis uses structured datasets with key factors such as job satisfaction, salary, career growth, and work-life balance. Models are evaluated based on key metrics such as accuracy, precision, recall, and F1-score to determine improvements in predictive performance after applying SMOTE. The results reveal how boosting techniques with SMOTE enhance the performance of various models, providing HR professionals with insights to design more effective, targeted retention strategies and reduce attrition risks.