基于GWR和sDNA模型的广州市路网形态对住宅价格影响的时空分析

Measuring Street Layout’s Spatio-Temporal Effects on Housing Price Based on GWR and s DNA Model: The Case Study of Guangzhou

基于地理加权回归（GWR）和扩展空间句法（sDNA）对传统的特征价格模型进行改进，进而提出基于路网的局部特征价格模型（NLHM），并对2014--2015年广州市路网形态对住宅价格的影响进行时空分析。研究发现：①不考虑路网变量影响时，地铁、物业管理费、是否精装修等变量对住宅价格的正向作用显著。②广州市中心城区主要位于接近度和穿行度的前景网络核心，城市郊区主要位于接近度和穿行度背景网络。③接近度与住宅价格呈正相关关系，穿行度与住宅价格呈负相关关系，接近度对住宅价格的影响大于穿行度对住宅价格的影响，二者对住宅价格的影响都随着计算半径的增加而减弱。④接近度系数空间格局具有跨尺度相似的特征，呈现出“郊区高，中心区低”的空间格局。穿行度系数空间格局具有出跨尺度变异的特征，中心城区在局部尺度上体现为穿行度系数绝对值的高值区，而在混合尺度和全局尺度上体现为低值区。⑤地铁变量系数随路网形态变量计算尺度的增加而减小，物业管理费、车位配比、是否精装修等变量系数变异程度较小。

In this paper, a network-based hedonic model （NLHM） is proposed based on geographically weighted regression （GWR） and extended space syntax （sDNA）, which improves the traditional hedonic price model. Using NLHM, we focus on the spatio-temporal effects of street layout on the housing price in Guangzhou from 2014-2015. Results provide the following conclusions. （1） If we don＇ t consider the effect of road network, variables like subway, property management fees and whether fine decoration or not will exert a significant positive effect on housing price. （2） The central area of Guangzhou is mainly located in the foreground network core of closeness and betweenness, while the suburban area is located in the background network core. （3） There is a positive correlation between closeness and hosing price, whereas a negative correlation is shown between betweenness and hosing price. The influence of closeness on hosing price is greater than that of betweenness, and the two parameters＇ influence coefficients will decrease if analysis radius increases. （4） The spatial patterns of closeness in different analysis radius illustrate a similar characteristic that closeness is high in suburban area and low in central area, while the spatial patterns of betweenness depict distinctive characteristics in different radius, for instance, the central area demonstrates high value and low value in local scale radius and global scale radius respectively. （5） The coefficient of subway variable decreases with the increase of analysis radius, while the variation of other variables＇ coefficients （e.g. property management fees, the ratio of parking, whether fine decoration or not） is smaller.