| Summary: | 碩士 === 國立成功大學 === 都市計劃學系碩博士班 === 96 === The research of hedonic price function in housing price don’t allow for spatial factor. In factor spatial factor has the influence on housing price model and amounts the question of spatial autoregression. Some several houses transaction material, it appear the attribute same or the characteristic same, but it presents the transaction price actually is different, it reason possibly is two positions is different, can have the different price, this reflected the people to the position by chance, form the position value also to have differently.
Because the house price mutually affects has the concept which the space dependence on one another, also is called "the spatial autoregression". Around the high house price also is the high house price gathers, but low house price then is opposite. This is the question of spatial autoregression.The residual mutually comes under the influence, the basic assumption of identical and independent distribution (iid) of the housing price variation would very possibly be violated.If may know gathers the place is located where, lead-in these spatial factors in the model, explains the spatial position with these spatial attribute variables, will be allowed the distinct improvement spatial autoregression question.
The aim of the paper combined these GAM , KRIGING method and GWR to treat the spatial autoregression question and to compare the difference of the three methods. The paper shows a good result in R square and accountable ability and find Space Cluster . The results also show that that these GAM , KRIGING method and Geographical weighted regression model indeed has improved the estimation accuracy of the court auction derached house price , compared to linear hedonic price model.
|
|---|
