아파트 가격 지수 산출에 관한 연구: 머신러닝 알고리즘을 중심으로

The adequacy and accuracy of the information provided by the apartment price index to policy decision makers is critical. However, the previously used apartment price index of the Korea Appraisal Board (KAB) has problems due to repeated selling models, including regional segmentation, non-reflection of external factors, and smoothing problems. These problems were also recognized in the preceding study, but still regional segmentation and external factors were not taken into account Therefore, in this study, the rate of housing loans, suicide rate, and the ratio of senior citizens were collected using the public data of the Seoul Metropolitan Government. In addition, various external variables, such as the number of nearby subway stations using spherical substrates, were used to divide them into regions and reflect them in the apartment price index. Next, the actual transaction price of 751,537 apartments was divided by training and test data at a ratio of 7:3, respectively, and the RMSE of the gradient boosting model (GBM) was the lowest at 0.143. Through the GBM index, it was confirmed that the factors behind the steady rise in apartment prices in some speculative areas, even in the midst of real estate regulation policies and financial crisis situations, are the reasons for the apartment’s external variables outside In addition, after diagnosing smoothing through the partial adjustment regression model, the KAB index was found to be larger than the GBM index. By doing so, it is expected that the use of the GBM index as an index for apartment prices could help establish real estate policies amid various financial situations.