Adding Climate Vulnerability to Real Estate Models: A comparative Machine Learning Analysis of Traditional Versus Climate-Enhanced Property Valuation

This project examines whether comprehensive climate vulnerability data improve real estate price predictions compared to traditional valuation models. While existing research primarily focuses on single climate hazards, such as flooding, I incorporated multi-hazard and baseline vulnerability data from the U.S. Climate Vulnerability Index to assess whether broader climate metrics enhance predictive accuracy. I developed and compared three machine learning models, including elastic net, random forest, and XGBoost. The project focused on three U.S. Cities – Milwaukee, Houston, and Denver. This allowed me to concentrate on neighborhood-level predictions, which provide the most benefit to individual homebuyers, local real estate professionals, and local municipalities. These cities allowed for comparison of locations with varying climate vulnerabilities and hazards.