顾及样本随机变异与网络约束的城市楼面地价时空分异规律研究

With the rapid urbanization, urban land price has kept increase and the bubble growth in real estate market becomes obvious. Reasonable spatiotemporal analysis model for urban land price is quite imperative to explore the distribution characteristics and driving factors of urban land price per floor area, so as to provide references for land market regulation and land resources configuration. Meanwhile, the development and popularization of GIS and Spatial Big Data provides new opportunities for the research of land price..In this research, we introduce a space-time analysis framework for urban land price per floor area based on Bayesian statistics and geographically weighted regression (GWR). It contains three aspects:.(1)To eliminate the statistical bias resulted from the sample random variation and spatial dependence, we introduce a Bayesian space-time model to estimate the spatial effect, time effect and spatial-time interaction effect of land price more precisely..(2)To make the traditional GWR more accurate when depict the non-stationarity in time and space of complex factors, we propose a mixed geographically and temporally weighted regression considering the network distance and parameter effect..(3)We illustrate our approaches in simulation experiments and make a case study in Shenzhen to explore the spatial-temporal pattern and factors of land price..It is expected that the research findings may supplement the method in field of geography and land management, and may be useful for the local land department to a better management and decision for urban land price.

快速城市化背景下，城市土地价格持续上涨，房地产市场泡沫化趋势明显，迫切需要建立合理的城市地价时空分析模型进行分布特征及其驱动因素的深入探讨，地理信息技术、空间大数据技术的发展与普及，为地价规律研究提供了新的思路。本课题在深入研究贝叶斯统计推断、地理加权回归模型的基础上，建立面向城市住宅用地楼面地价的贝叶斯时空分析模型，准确评估土地价格的空间效应、时间效应及空交互效应，消除地价样本随机变异与空间依赖导致的随机噪声；并构建顾及道路网络距离的混合时空加权回归模型构建时空权重矩阵，提高传统地理加权回归模型在刻画复杂要素的时间、空间非平稳性的精度，充分考虑了网络距离、变参数与产参数的效应；在此基础上分别进行模拟实验和典型城市（深圳）应用分析，实现对城市楼面地价时空分布及其充分驱动因素的估计，其研究成果不仅能够丰富地理学和土地管理领域的理论与方法，为精细化土地价格管理与评估提供技术支撑。

快速城市化背景下的城市土地价格持续上涨，房地产市场泡沫化趋势明显，迫切需要建立合理的城市地价时空分析模型进行分布特征及其驱动因素的深入探讨，地理信息技术、空间大数据技术的发展与普及，为地价规律研究提供了新的思路。本课题在深入研究贝叶斯统计推断、地理加权回归模型的基础上，建立面向城市土地价格的贝叶斯时空分析模型，准确评估土地价格的空间效应、时间效应及空交互效应，消除地价样本随机变异与空间依赖导致的随机噪声；并构建顾及道路网络距离的混合时空加权回归模型构建时空权重矩阵，提高传统地理加权回归模型在刻画复杂要素的时间、空间非平稳性的精度，充分考虑了网络距离、变参数与产参数的效应；在此基础上分别进行模拟实验和典型城市（深圳）应用分析，实现对城市住宅用地楼面地价时空分布及其充分驱动因素的精准估计，已发表各类学术论文十篇。