Machine Learning Approaches to Salary Prediction in Human Resource Payroll Systems

Abstract

Salary prediction plays a vital role in human resource (HR) management, enabling organizations to streamline payroll systems, improve decision-making, and ensure fair compensation. Accurate salary forecasting supports workforce planning, budgeting, and employee retention strategies. Traditional payroll systems often rely on static rules and historical records, which may not capture complex relationships between employee attributes and income levels. With the advent of machine learning (ML), predictive models have emerged as powerful tools for addressing these limitations in HR payroll systems. This study proposes an Extreme Gradient Boosting (XGBoost) model for salary prediction using the Adult Income Dataset. The method incorporates feature selection following data pretreatment, which includes managing missing values, eliminating outliers, one-hot encoding, and min–max normalization, in order to maintain the most relevant characteristics. To guarantee accurate model assessment, the dataset is separated into training (80%) and testing (20%) subsets. With an AUC-ROC of 0.93 and 91.16% accuracy, precision, recall, and F1-score all at 88%, the suggested XGBoost model demonstrated high predictive performance.  The findings show that the XGBoost model performs noticeably better than more conventional models, such as Naïve Bayes (NB) and Support Vector Machine (SVM), making it a dependable and efficient method for predicting salaries in payroll systems. This study highlights the potential of advanced ML techniques to enhance efficiency and accuracy in HR management.