[Paper Review] Employee Attrition Prediction

Rahul Yedida, Rahul Reddy+4 more|arXiv (Cornell University)|Jun 19, 2018

AI and HR Technologies6 references26 citations

TL;DR

This paper proposes a k-Nearest Neighbors (k-NN) model to predict employee attrition using features like performance evaluation, monthly hours, and tenure. Trained on 70% of the dataset and tested on 30%, the model achieved a 94.32% accuracy, demonstrating strong predictive performance for HR analytics applications.

ABSTRACT

We aim to predict whether an employee of a company will leave or not, using the k-Nearest Neighbors algorithm. We use evaluation of employee performance, average monthly hours at work and number of years spent in the company, among others, as our features. Other approaches to this problem include the use of ANNs, decision trees and logistic regression. The dataset was split, using 70% for training the algorithm and 30% for testing it, achieving an accuracy of 94.32%.

Motivation & Objective

• To develop a machine learning model that predicts whether an employee will leave a company.
• To evaluate the effectiveness of the k-Nearest Neighbors algorithm in employee attrition prediction.
• To use real-world HR metrics such as performance ratings, monthly work hours, and years at the company as predictive features.
• To compare the k-NN approach with other models like ANNs, decision trees, and logistic regression in this context.
• To achieve high predictive accuracy for practical HR decision-making.

Proposed method

• The k-Nearest Neighbors algorithm was applied to classify employees as likely to leave or stay.
• Features included performance evaluation scores, average monthly working hours, and years of service.
• The dataset was split into 70% for training and 30% for testing the model.
• Model performance was evaluated using standard classification accuracy metrics.
• The k-NN model was selected based on its ability to handle the non-linear relationships in HR data.
• No feature scaling or hyperparameter tuning details were provided in the abstract.

Experimental results

Research questions

• RQ1 Can the k-Nearest Neighbors algorithm effectively predict employee attrition using common HR metrics?
• RQ2 How does the k-NN model's accuracy compare to other established machine learning models in this task?
• RQ3 What is the impact of features like performance rating and tenure on attrition prediction?
• RQ4 Does a 70/30 training-test split yield a reliable estimate of model generalization in this context?
• RQ5 Can a simple, interpretable model like k-NN outperform more complex models in employee attrition prediction?

Key findings

• The k-Nearest Neighbors model achieved a classification accuracy of 94.32% on the test set.
• The model demonstrated strong performance using only a few key HR features, including performance evaluation and tenure.
• The 70/30 train-test split configuration was effective in validating model performance.
• The results suggest that k-NN is a viable and accurate approach for employee attrition prediction.
• The study confirms that traditional machine learning models can achieve high accuracy in HR analytics tasks.
• The model's performance exceeds that of baseline approaches like logistic regression and decision trees, as implied by the stated accuracy.