SalaryScale AI: An Intelligent and Explainable Machine Learning System for Salary Prediction

Abstract

Accurate salary prediction is essential for supporting fair compensation planning, recruitment, and career decision-making. However, predicting salaries remains challenging due to the complex relationships among professional experience, technical skills, education, and geographical factors. This study presents SalaryScale AI, an intelligent and explainable machine learning framework developed using the Stack Overflow Developer Survey dataset to provide accurate and transparent salary predictions. The proposed system employs a stacking ensemble of XGBoost and LightGBM models, optimized through Optuna-based Bayesian hyperparameter optimization with GPU acceleration to improve predictive performance. Extensive data preprocessing and multi-label feature engineering were performed to effectively handle high-dimensional survey data. The proposed model achieved an R² score of 0.61, outperforming conventional regression approaches while improving overall prediction accuracy. To enhance model interpretability, SHAP (SHapley Additive exPlanations) was integrated to explain both global and individual salary predictions. In addition, Evidently AI was incorporated to monitor data drift and support long-term model reliability. The trained model was deployed using a FastAPI-based architecture, enabling low-latency, real-time salary prediction through RESTful application programming interfaces. The experimental results demonstrate that the proposed framework provides an accurate, scalable, and interpretable solution for salary prediction, making it suitable for practical human resource analytics and intelligent decision-support applications.