How the Pattern Recognition Ability of Deep Learning Enhances Housing Price Estimation

• Published in: 한국경제지리학회지, 25(1) pp. 183 - 201
• Main Authors: 김진석, 김경민
• Format: Journal Article
• Language: English
• Published: 한국경제지리학회 01.03.2022
• Subjects: 지리학
• ISSN: 1226-8968; 2713-9115
• DOI: 10.23841/egsk.2022.25.1.183

Estimating the implicit value of housing assets is a very important task for participants in the housing market. Until now, such estimations were usually carried out using multiple regression analysis based on the inherent characteristics of the estate. However, in this paper, we examine the estimation capabilities of the Artificial Neural Network(ANN) and its ‘Deep Learning’ faculty. To make use of the strength of the neural network model, which allows the recognition of patterns in data by modeling non-linear and complex relationships between variables, this study utilizes geographic coordinates (i.e. longitudinal/latitudinal points) as the locational factor of housing prices. Specifically, we built a dataset including structural and spatiotemporal factors based on the hedonic price model and compared the estimation performance of the models with and without geographic coordinate variables. The results show that high estimation performance can be achieved in ANN by explaining the spatial effect on housing prices through the geographic location. KCI Citation Count: 0