Machine Learning and House Price Prediction

While house prices seem to be strongly determined by neighborhood and bedroom count, there are factors that contribute to a house’s sale price that may not be initially considered. House price prediction modeling has gained popularity as of late in an effort to increase transparency within the industry. This project creates two models which can predict the housing price for homes in Ames, Iowa, using a simple linear regression, and a random forest model. Using house features (house condition, dwelling type, zoning, etc.) and their respective final house prices, this project was successful in developing a model with predictive power. Ultimately, the random forest model, the more sophisticated ensemble method, yields a better prediction with a final RMSE lower than that of the simple regression. Finally, a clustering method allows homes to be divided into ten clusters based on a k-means cluster of six chosen features.