Interpretable House Price Prediction Using Ensemble Learning and Explainable AI Techniques

This study presents a ensemble learning based machine learning approach for predicting house prices on a structured tabular dataset. Within the scope of the study, four regression models are investigated, which are Bayesian Ridge Regression, Elastic Net Regression, Stochastic Gradient Descent (SGD) Regression and a Voting Regression model combining the previous three models. Experimental results demonstrate that Bayesian Ridge and SGD Regression achieved the highest predictive accuracy, while the Voting Regression model provides robust yet slightly lower performance. Additionally two Explainable Artificial Intelligence (XAI) techniques, SHAP and LIME, are employed to interpret the Voting Regression model’s predictions, revealing that Square_Footage is the most influential feature, followed by Lot_Size and Year_Built features and with minimal contributions from Neighborhood_Quality feature. The alignment between global and local interpretability methods further validates the model’s decision-making transparency and robustness. Consequently, this study indicates the effectiveness of combining traditional regression models with ensemble strategies and explainability techniques for building accurate and interpretable predictive systems for real estate price estimation, which provides a foundation for future research into incorporating broader contextual and market-driven features.