Determining home listing price with the use of machine learning and neural networks

This study tries to predict the home price at the time of listing for sales. Home price prediction is important for the owner, real estate agents, financial institutions, and government agencies. Mass appraisal with the use of machine learning is an evolving field. It saves time, money, and effort and we can do multiple appraisals at the same time. Data from a real estate agent from rural Wisconsin counties are used. Three methods are applied for home price prediction. Random forest, extreme gradient boosting, and neural networks. All methods were chosen from the review of the literature and previous studies. The most useful method found was extreme gradient boosting. Random forest came a close second. I found neural networks not much useful. This finding was due to a better fit of random forest decision trees for the type of data. The data was categorical and numerical with many columns converted to one-hot encoding. Random forest works well when multiple complex situations exist. This paper shows why random forest and extreme gradient boosting would be useful for home price prediction.