Employee Performance Prediction in Human Resource Management Using Gradient Boosting Machines (GBM)

Abstract

The issue of predicting employee performance has become an important one in contemporary Human Resource Management (HRM), as businesses try to use the data analytics approach for making evidence-based decisions in such areas as talent acquisition and retention. However, the existing approaches that are used by companies to evaluate employees' performance involve subjective assessments and periodic reviews, which do not provide accurate results and waste valuable company resources. In this paper, an ML model is introduced that is based on the algorithm of GBM and has high prediction accuracy related to organizational performance within the scope of HR. Specifically, in this research, an ML model including the preprocessing pipeline with imputation, encoding, and Min-Max Normalization of HR features and SHAP analysis of HR feature importance is introduced. The proposed solution has been built on the basis of the publicly available IBM HR Analytics Employee Attrition and Performance data set with 1,470 samples and 35 features. For the sake of improving the GBM classifier. From these experiments, it is evident that the accuracy rate for GBM model could reach 94.2%, precision 98.8%, recall 95.6%, F1 measure 97.2%, and AUC-ROC value of 0.976, which is notably superior compared to conventional classification algorithms, like Decision Tree (93.55%), Random Forest (92.57%), Support Vector Machine (SVM) (85.30%), and Logistic Regression (67.54%). Further validation of the effectiveness of the processes employed through feature engineering and hyperparameter tuning is shown in an ablation experiment, which demonstrated that the accuracy of predictions increased after going through those processes. These experiments highlight the strength of GBM models compared to others in balancing class distribution, non-linear relationships among features, and high-dimensional HR data. This approach can readily be applied with enterprise HR software solutions for enhanced talent management and organizational productivity. Future work will be centered on incorporating the deep learning and real-time prediction functionality of the system.