Deep Neural Network As a Tool for Appraising Housing Prices: A Case Study of Busan, South Korea

Abstract

This study examines whether the number of hidden layers in a deep neural network significantly influences the model accuracy and efficiency for appraising housing prices. We provide empirical evidence that the deep neural network can achieve high accuracy with a small number of hidden layers on our dataset, which contains various hedonic variables. Furthermore, we show that adding layers does not necessarily guarantee the model’s accuracy and effectiveness of the computing time.