A Systematic Literature Review of Quantitative Models for Predicting Employee Attrition

This study aims to address the significant employee attrition problem by conducting a thorough and organized analysis of the existing literature on machine learning (ML) approaches for predicting attrition in different sectors. The analysis includes a comprehensive review of more than thirty research publications, focusing on the data types, characteristics studied, algorithmic methods, and measures used to assess predicted effectiveness. The assessment acknowledges the IBM HR dataset as a commonly used benchmark. It acknowledges several aspects that impact attrition, such as the work environment, job satisfaction, salary, and possibilities for professional progress. Conventional algorithms, including Random Forest, Logistic Regression, Support Vector Machine, and Decision Trees, have been extensively used, with a noticeable shift towards ensemble approaches to improve accuracy and resilience. The study highlights the need to handle imbalances in datasets, with widely used techniques such as ADASYN and SMOTE employed to enhance models' dependability. This study presents the current status of machine learning (ML) methods in predicting employee turnover. It emphasizes the need to have access to real-life datasets and expand research to include ensemble techniques. The study's results indicate progress towards more knowledgeable and strategic methods in human resource management.