Prediction of Employee Attrition Classification and Retention Strategies Using Gradient Boosting with Binary Logistic Regression

A voluntary or involuntary parting of an employee from an organization is known as and the strategies to maintain an employee are known as Employee attrition (turnover) and retention. Employee attrition and retention are crucial Human Resource (HR) concerns, where attrition impacts the organizations performance, profitability, and productivity. Several methods have been proposed to predict attrition of employees, but the existing methods have overfitting, class imbalance, interpretability, and generalizability problems. To overcome these issues, a novel Gradient Boosting-Binary Logistic Regression approach is proposed to predict the employee's turnover in an early stage. Initially, an IBM-HR dataset is considered as the input for preprocessing, which is performed using Standard Scalar to preserve data distributions. Next, One-hot encoding is assessed for data encoding to convert categorical variable into numerical variables. Followed by, feature extraction using Principal Component Analysis (PCA) for faster training and enhancing the performance of model. Later, Gradient Boosting-Binary Logistic Regression (GB-BLR) is employed for classification to analyze and predict attrition of employee's. The proposed GB-BLR method obtained better performance when compared with other existing Extra Trees Classifier (ETC) method in terms of accuracy, precision, recall, and f1 score of 97.48%, 98.57%, 96.25%, and 98.77% respectively.