Efficient Approach to Employee Attrition Prediction by Handling Class Imbalance

The key success of any organization lies on its employees and thus ability to monitor employee attrition efficiently becomes important. Our work aims to use machine learning models to accurately predict on whether an employee would decide to leave a company or not. Main algorithms used for this purpose include Logistic Regression, KNN and Weighted Decision tree. Another aspect that needs to be looked into is the amount of classification imbalance present in this problem statement. Our work has applied multiple data based and algorithmic approaches to resolve this imbalance. The use of SMOTE, Near-Miss, ADASYNC and other techniques provide an efficient and balanced prediction of employee attrition. From all the experiments we have done, we arrive at a proposed hybrid model that combines data based and algorithmic method to resolve imbalance that accurately predicts attrition of up to 98.4%, proves recall rate of 98.43%, shows precision of 97.69% and an excellent F1 score of 99.17% . Our analysis would help companies understand the important areas that needs to be focused on to avoid attrition and hence lead to growth and higher achievements.