Forecasting Short-Term Housing Transaction Volumes using Time-Series and Internet Search Queries

초록

To understand the current status of the housing market, a short-term predictor of housing transaction volume, which is one of the key indicators of the housing market, is important. However, short-term fluctuations in the volume of housing transactions make short-term forecasting difficult. For the short-term forecasting of housing transaction volumes, this study developed a hybrid model based on both Auto-Regressive Integrated Moving Average (ARIMA) analysis and regression analysis using Internet search frequency. In particular, the model focused on forecasting the trading volume of apartments, which constitute the greatest number of residential types in Korea. In this study, the mean average percent errors (MAPE) of the short-term prediction of a ARIMA model and a hybrid model were compared. As a result, the MAPE of the hybrid model was improved by about 50% (6% p) compared with the ARIMA model. It is expected that the proposed hybrid model will be used in national policy-making for the housing market and will be the basis of a study on the forecasting of housing transaction volume.

키워드