A Supervised Machine Learning Model to Optimize Human Resources Analytics for Employee Churn Prediction

Employee churn is one of the most daunting challenges that an organization is likely to face in its lifecycle. An unexpected employee departure can adversely impact service delivery, reduce productivity and customer loyalty. It is therefore pertinent to predict employee churn to help organizations retain valuable employees. This paper proposes a model that makes use of feature selection through Pearson Correlation Method, Information Gain and the Recursive Feature Elimination together with robust classification methods that include Random Forest (RF), Logistic Regression (LR), Decision Trees (DT), Gradient Boosting Machines (GBM) and K Nearest Neighbors (KNN) to predict employee churn. The training and testing data were obtained from the IBM dataset. The accuracy of the algorithms improved after applying the feature selection methods. Experimental results showed that Random Forest performed better than all the comparative algorithms in terms of classification accuracy. Consequently, the algorithm demonstrated to be a more appropriate algorithm in predicting employee churn.