Using Artificial Intelligence-Based Machine Learning Regression Models for Predictions of Home Prices

Abstract

Accurate forecast of home prices is crucial in making informed decisions in real estate, financial plans and investment practices. This work exploits the Volusia County Property Sales Dataset to create a strong machine learning model for predicting home prices. By undertaking extensive data preprocessing, feature engineering, and transformation of the outcome and having an XGBoost regression model as a result, we trained and measured its performance Metrics include R², MSE, and MAE.  The XGBoost model outperformed baseline models such as Random Forest (R² = 0.87) and Artificial Neural Networks (ANN) with R² = 0.97, MSE - 8.17, and MAE = 4.03, R² = 0.84. These results show the effectiveness and accuracy of XGBoost when identifying complex, nonlinear relations in the housing data. The proposed pipeline is scalable and can be adapted to fit into live property valuation systems, lending valuable insights to stakeholders in urban planning, real estate investments, and public policy formulations.