# 教師資料查詢 | 類別: 期刊論文 | 教師: 葉怡成 YEH, I-CHENG(瀏覽個人網頁)

The purpose of this paper is to propose a new method called stepwise decomposition regression analysis method to overcome the shortcomings of traditional multivariate regression analysis in easy understanding, versatility, application, model elasticity and so on. The principle is to assume that the price per unit area of real estate is the average price per unit area of the specific circle of housing supply and demand multiplied by the product of several dimensionless adjustment coefficients of factors. This method starts from the most important factor, then one by one, to decompose the estimated adjustment coefficient of each factor, and build the predictive model for each adjustment coefficient. This method consists of three steps: (1) Sorting: Employ sorting and grouping approach to estimate the importance of the factors. (2) Decomposition: Use the stepwise decomposition approach to construct the single variable regression model for each adjustment coefficients to its factor. (3) Integration: Integrate the adjustment coefficient regression models to a real estate price valuation model. The dependent variable in this study is the residential housing price per unit area. The independent variables include the distance to the nearest MRT station which represents the impact of transportation function, the number of convenience stores in the living circle on foot which represents the impact of living function in the living circle on foot, the age of house which represents the impact of living function in room, the transaction date which represents the impact of market trend, and the geographic coordinates which represent the impact of spatial location. The samples are collected from two districts in Taipei City, and two districts in New Taipei City, totally four circles of supply and demand, and are divided into four data sets. The results show that the stepwise decomposition regression analysis is better in easy understanding, versatility, application, model flexibility, and reach a higher degree of accuracy than conventional multivariate regression analysis.

ISSN2219-0953