국지가중회귀(Locally Weighted Regression) 모형을 활용한 아파트 실거래가격지수 산정방법에 관한 연구

A real estate price index tracks the price of a standard unit of housing over time. The index typically is constructed from observations on the sale price and their characteristics. Due to the requirement of the stability of the index, I assumed four different types of houses. With centering on them, I estimated locally weighted regressions (LWR) of hedonic price models and constructed the transaction-based price indices from January 2006 to October 2010 for Gangnam Gu, Seoul. The appraisal-based KB indices have appraisal smoothing problems in which they have low volatility and time-lag but the proposed transaction-based price indices do not show those problems.