Prediction of Employee Attrition Using Ensemble Model Based on Machine Learning Algorithms

Author

Zakwa, Terrence

Title

Prediction of Employee Attrition Using Ensemble Model Based on Machine Learning Algorithms

Abstract

Employee attrition denotes the nonstop drop in the number of workers in an organization by the process of withdrawal, abdication, or death (Dutta et al,. 2020). Employee attrition is expressed as the normal process by which the employees leave the organization due to some reasons, such as the resignation of employees. There are many factors that can cause employee attrition (Peng, 2021). The employees leave the organization faster than they are hired. When the employee leaves the organization, the vacancies remain unfilled, resulting in a loss for the organization. The first objective is to analyse different machine learning algorithms used in HR datasets to predict attrition. The second objective is to design and implement a machine learning model based on Ensemble Modelling which predicts employee attrition using artificial neural network. The final and last objective is to evaluate the effectiveness of the Ensemble Modelling method and machine learning in employee attrition prediction.Therefore, to this end, the researcher managed to develop a system or model that uses ensemble modelling for various ML algorithms which are Logistic Regression, SVM, Linear SVC, KNN, Naive Bayes, Decision Tree, Gradient Boosting Trees and Random Forest algorithm to predict which employee is likely to leave using a dataset of an unknown organization found on Kaggle. All the machine learning algorithms were combined using ensemble modelling and the results were satisfying. The employee attrition rate helps to understand the progress level of an organization. The researcher performed all the necessary black, white box tests and performance tests using the confusion matrix, the author found that the system had satisfactory performance. The system was tested in accuracy, misclassification error/error rate and it achieved 91.2% and 0.8% respectively. The model attained an overall precision of 86% and a sensitivity or recall of 84%. An F1 score of 91.3% was achieved with a specificity or true negative rate of 96%.

Date

2023

Publisher

BUSE

Keywords

Deep Learning

Ensemble Modelling

Employee Attrition

Supervisor

N/A