Gender-Based Analysis of Employee Attrition Prediction Using Machine Learning

Abstract

Employee turnover is a significant problem in organizations because it comes with productivity and cost implications. This paper focuses on predicting employee turnover using machine learning techniques that incorporate gender aspects. We used strong Random Forest classifiers to predict attrition based on a wide cross-section of the employees’ activities and the feature importance assessment. The procedure involved data cleaning, splitting the dataset for males and females, creating models for them, and using assessment tests with different measures. When we separated the data base by gender, our analysis identified unique factors that predisposed the two groups to dropping out. The importance of features, the ROC curve, and the SHAP map showed how variables such as "job role," "monthly income," and "work-life balance" affected attrition differently between males and females. For female employees, job satisfaction and time directly influenced attrition, whereas for male employees, previous companies and distance from home had a greater impact. The results of the research therefore imply the need for gender-sensitive HR practices that can inform the development of gender-sensitive accommodation policies as a way of responding to the challenges facing each gender. This approach aids in the explanation of attrition tendencies and the provision of better organizational practices.