Home price prediction using enhanced features from the environment

Accurate home sale price prediction is essential for stakeholders such as homeowners, real estate investors, banks, and policymakers. This study enhances predictive accuracy by integrating advanced feature engineering with traditional structural attributes. Using datasets from Redfin and Open Street Map, we develop a predictive model for home sale prices in Washington DC, Arlington VA, and Alexandria VA. The methodology begins with a baseline model relying solely on structural features, such as square footage, number of bedrooms, and home age. We then enhance the model by integrating environmental features derived from Open Street Map, including proximity to green spaces, public transportation, and educational facilities. Subsequently, we further enrich the model with additional variables, such as the total greenspace area within specific thresholds around each home, and whether a highway needs to be crossed to access the largest greenspace within a certain distance from each home. At each stage, we build a new model trained on 80% of the initial data and evaluate its performance on a test set comprising the remaining 20% of the data, using R-squared as the evaluation metric. The final model achieves an R-squared score of 0.868, a significant improvement over the baseline model’s R-squared score of 0.715, demonstrating that incorporating detailed environmental features substantially enhances home price prediction accuracy.