Predicting Employee Attrition Using Behavioural Data

Abstract

Employee attrition is a massive problem faced by companies and organisations. It increases the cost of recruiting employees, the continuity of work is disrupted and productivity is lowered. With a vast volume of employee datasets and HR specifics, Machine Learning is used to predict employee turnover. This helps the authorities to plan retention strategies and take precautionary measures to prevent employee turnover. This paper proposes the development of end-to-end solutions related to predicting the turnover of employees based on the availability of synthetic data. This model comprises of generation of data, processing of data, exploratory data analysis, training of models, evaluation of models, selection of models, and interpretability of models. The models used for training are Decision Tree, AdaBoost, XGBoost, Support Vector Machine, Gradient Boosting, Logistic Regression, Random Forest, and K-Nearest Neighbours. The models are analysed using standard parameters of precision, accuracy, recall rate, and F1 score values. A web interface is developed that collects the data and shows the results. The result contains the prediction table and SHAP values, which help explain the prediction.