Ensemble Based Attrition Prediction in Corporate Settings

The phenomenon of employees leaving a business, or employee attrition, is a serious problem with broad societal consequences. High attrition rates cause organizational stability to be disrupted, which has an effect on economies and communities by causing the loss of institutional knowledge and valuable personnel. This study uses a comprehensive method to forecast employee attrition that combines deep learning and machine learning models. Recursive Feature Elimination (RFE), Principal Component Analysis (PCA), and Non-Negative Matrix Factorization (NMF) are used methodically to choose features. SMOTE-ENN and Borderline SMOTE oversampling techniques are used to address imbalanced datasets. XGBoost, CATBoost, and a Gated Recurrent Units (GRU) model are used for attrition prediction to combine their predictive power; the GRU model uses temporal information to improve accuracy. Moreover, to improve model accuracy, optimization methods like Grid search optimization and Bayesian optimization are used. The proposed system gives an accuracy of 99%. The potential of machine learning and deep learning in addressing real-world challenges is highlighted by the combination of sophisticated feature selection, efficient oversampling, optimization techniques, and cutting-edge models.