Predicting Home Value in California, United States via Machine Learning Modeling

The market value of real estate is difficult to predict with simple regression model due to the diversity and complexity of the real data. In this paper, with the latest real estate data of three counties in Los Angeles, California, United States, both linear and non-linear machine learning methods are employed to predict the log error of the home value. The motivation is to improve the accuracy in home value prediction with advanced methods. The main contribution is that it finds that traditional linear models are not predictive for complex home value data sets, while tree based non-linear models are most accurate with the lowest mean square errors.