Machine learning and residential real estate: Three applications

Abstract

This thesis is a composite of empirical studies for three topics, automated valuation model (AVM), residential property price index (RPPI), and the analysis of land development. Firstly, machine learning techniques are applied to develop the implementations of AVMs, whose purpose is to provide a price estimate of a particular property at a specified time. The main objective is to minimize human intervention in price estimation when the presence of missing values remains a major challenge in the process. Then, the proposed AVM implementation is applied to compiling the residential property price index, which tracks the trend of market values, cooperating with the classic indexing approaches. The main objective is to investigate whether more accurate price predictions lead to a better price index and examine how well the machine learning techniques explain the time effects. Thirdly, land development is a "real option" that allows the landowner to decide whether and when to develop the vacant land by spending money. The analysis of land development is to examine the real option, including the valuation of the option and the optimal timing to exercise the option. The research uses machine learning techniques with the factors on both the investment output (residential buildings) side and the investment cost (construction cost) side, such as the growths and uncertainties of property prices and construction costs.