기계학습 기반 주택시장의 조기경보체계*

Abstract

This study proposes an early warning system for risks of the housing market based on machine learning models. We adopt a signal approach to detect the housing market risk and establish the early warning system using classification methods. Considering the moment when the housing market falls into recession as a warning signal, we set the signal as the price which is more than the sum of the average and standard deviation of upcoming prices. The detected signals are consistent with empirical observations in the Korean housing market. We select the best performing function among classification models for machine learning which predicts a warning signal. As a result of an intercomparison of models including the logistic regression, the support vector machine, the random forest and the artificial neural network with the use of inputs such as housing price indices, macroeconomic variables and other housing market variables, we find that the random forest demonstrates the highest prediction performance. Our early warning system yields policy implications in terms of relevant detection of price fluctuations in the housing market.