基于GWR模型的南京主城区住宅地价空间异质性驱动因素研究

Driving Factors on the Spatial Heterogeneity of Residential Land Price in Downtown Nanjing Based on GWR Model

以南京市主城区为例,基于2014年住宅用地监测点数据,利用GWR模型分析各驱动因素对住宅地价贡献程度及空间异质性成因,以促进地价的科学规范化管理。研究表明:①各影响因素对住宅地价的平均边际贡献程度为:CBD>河流水系>城市快速路>大学>医院>公园绿地>小学>幼儿园;②CBD仍占据影响住宅地价的主导地位,城市公共设施对住宅地价的负向效应有扩大趋势,"教育型"设施逐渐成为居民购房行为的重要影响因素,面状自然地物在空间影响模式上存在"互补效应";③城市住宅地价影响模式由单一影响因素为主导向多因素共同作用转化,各影响因素虽作用大小仍有所侧重但差异逐渐缩小;④较全局OLS模型(50%),GWR模型可解释监测范围内87%的住宅地价变化,能够更加精准地研究土地市场的空间异质性。

This paper,taking the downtown of Nanjing as an research area and based on the final retained data from residential land monitoring sites in 2014,applies the geographically weighted regression(GWR)model to simulate the contribution of driving factors and the causes of spatial heterogeneity of residential land market in study area,so as to promote the scientific and standardized management of land price.The results indicate that:1)In the GWR model,the order of the average marginal contribution on the land premium from high to low is the distance from CBD,river,expressway,college,hospital,park,primary school and kindergarten.2)The distance from CBD has occupied the dominant position in influencing factors of residential land prices.The negative effect of urban public facilities on residential land prices tends to expand and the’educational-type’infrastructures have gradually become an essential influencing factor in the purchase behavior of residents.There is a’complementary effect’on the spatial influence model of surface natural features.3)The impact model of urban residential land price was dominated by a single influencing factor and then transforms to multi-factor interaction.The function intensity difference of each of influencing factors is gradually narrowing.4)The GWR model can be well applied to stimulate spatial heterogeneity of land market in target area accurately.The model could explain 87%of the price changes of residential land within the monitoring range.The interpreting abilities have been improved significantly than that based on global regression model(50%).