GWR모형과 GIS를 이용한 주택가격 추정에 관한 연구

Abstract

In this study, housing price was estimated based on apartment area, number of apartment floor, population density, straight line distance to city hall, subway, 2-lanes, 4lanes and park as explanatory variable by setting actual transaction value of Seoul city apartments as dependent variable using OLS, SAR, SEM (spatial econometrics model) and GWR model (geographically weighted regression model) that is a non-parametric estimation method. As a result of analysis, considering that in case of R2 value and log likelihood, GWR model was higher than OLS but in terms of reference value of both AIC and SC value, GWR model was represented to be lower than OLS, housing price predictability of GWR was proved to be the more reliable than that of OLS model. In addition, as a result of comparing residual distribution, it could be confirmed that while in OLS model, regional spatial correlation was prominent, spatial dependence was considerably moderated in GWR model. As a result of this study, as influence of housing by each features was confirmed to be differently represented spatially, at the time of establishing housing policy, authorities concerned of the government is required to establish and execute regionally differentiated customized housing policy rather than a uniformed policy based on sufficient understanding for the housing sub-market.