Machine Learning Approach for Predicting Employee Attrition and Factors Leading to Attrition

Abstract

In this study that aims to prevent the attrition of human resource which is so important for enterprises, as well as to prevent the leave of employment which is the natural result of such attrition, employee attrition and factors causing attrition are tried to be determined by predictive analytics approaches. The sample dataset which contains 30 different attributes of 1470 employees was obtained for the analysis from a database provided by IBM Watson Analytics. In the study, seven different machine learning algorithms were used to evaluate the prediction achievements. The gain ratio approach was preferred in determining the factors causing attrition. The key point of the study was to cope with the imbalanced data through resampling with bootstrapping. Thereby, even in the blind test, prospering prediction performances reaching up to 80% accuracy were achieved in robust specificity without sacrificing sensitivity. Therewithal, the effective factors causing attrition were investigated in the study and it was concluded that the first 20 attributes ranked according to their gain ratio were sufficient in explaining attrition.