Implementing machine learning for predictive analytics: An empirical study of employee turnover

Aug 5, 2026

Employee turnover presents a critical challenge for organizations, impacting operational efficiency, morale, and long-term performance. This study evaluates machine learning models for predicting employee attrition using the IBM HR Analytics dataset, comparing Logistic Regression, Decision Tree, Random Forest, Gradient Boosting, and Support Vector Classifier. Performance is assessed with accuracy, precision, recall, F1-score, ROC AUC, and cross-validation, where SVC achieves the strongest results (AUC 0.8129; cross-validated AUC 0.8291). Feature importance highlights overtime, job satisfaction, and employee involvement as key predictors, supporting proactive HR decision-making and future dashboard deployment.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages20

Reading Time50 min

AccessFree to download

Introduction

Problem and motivation

Turnover types and theoretical background

Machine learning approach and objectives

NEXRES 2 (2025) 100873 Contents lists available at ScienceDirect Next Research journal [homepage:](homepage: www.elsevier.com/locate/nexres)[ www.elsevier.com/locate/nexres](homepage: www.elsevier.com/locate/nexres) | Implementing machine learning for predictive analytics: An empirical study of employee turnover
*
Ghita Regasse , Francesco Venier
MIB Trieste School of Management, Italy | | | | --- | --- | --- | | A R T I C L E I N F O | | A B S T R A C T | | Key words:
Attrition prediction Employee turnover Predictive analytics Support vector classifier | | Employee turnover presents a critical challenge for organizations, affecting operational efficiency, morale, and long-term performance. This paper investigates the application of machine learning models for predicting employee attrition using the IBM HR Analytics dataset. Several algorithms were tested, including Logistic Regression, Decision Tree, Random Forest, Gradient Boosting, and Support Vector Classifier (SVC). Models were evaluated using accuracy, precision, recall, F1-score, ROC AUC, and cross-validation. Among them, Support Vector Classifier (SVC) model demonstrated superior performance, with an AUC of 0.8129 and cross-validated AUC of 0.8291, indicating strong discriminative capability. Feature importance analysis revealed that overtime, job satisfaction, and employee involvement are key predictors of turnover. The findings highlight machine learning’s value in transitioning HR practices from reactive to proactive workforce management, with potential future applications including deployment in HR dashboards, ethical compliance, and dynamic data integration. | 1. Introduction Employee turnover is a persistent and costly challenge for organizations across industries. The decision that an employee could make in order to decide either to stay within the organization or maybe leave is often impacted by a complex interplay of factors. High turnover rates not only affect an organization’s operational efficiency but also its financial stability, how the employee think, and overall performance. It is known that the cost causes by recruiting, and its related activities, coupled with the loss of institutional knowledge, can have a real influence on the company’s ability to maintain a competitive edge. As such, understanding and mitigating the drivers of employee turnover has become a critical area of focus for HR departments and organizational leaders. Despite the growing body of research on employee turnover, organizations still struggle to predict and prevent it. Traditional approaches to managing turnover, such as exit interviews and employee surveys, provide valuable insights but are often retrospective and reactive. These methods typically focus on understanding why employees leave after they have already made the decision to do so. This reactive approach, while helpful for improving HR policies and workplace culture, fails to address the core challenge: predicting and preventing turnover before it occurs. This brings us to the central problem this paper seeks to address: How can organizations predict employee turnover in real time and proactively intervene to retain top talent? The need for a predictive approach of turnover management has led to the exploration of “machine learning techniques”, which offer the possibility to analyze large datasets and be able to check the patterns that cannot be immediately visible through old HR practices. The aim here is to identify the best machine learning model that is suitable for deployment to predict employee turnover, providing HR professionals with actionable insights that enable timely intervention and retention strategies. “Employee turnover” could be split into two categories: voluntary and involuntary turnover. Voluntary turnover basically happens when employees choose to quit the organization, often for reasons such as job dissatisfaction, or better career opportunities. Involuntary t

FAQs

Which machine learning algorithms are compared for attrition prediction?

The study tests Logistic Regression, Decision Tree, Random Forest, Gradient Boosting, and Support Vector Classifier (SVC) using the IBM HR Analytics dataset.

How is model performance evaluated in the paper?

Models are assessed using accuracy, precision, recall, F1-score, ROC AUC, and cross-validation.

What factors are identified as key predictors of employee turnover?

Feature importance analysis shows overtime, job satisfaction, and employee involvement as key predictors of turnover.

Aug 5, 2026

Upload by Valentina