基于机器学习方法的城市社区尺度商铺租金空间布局分析

研究目的：通过引入机器学习方法，利用地理时空大数据研究精细尺度的商铺租金分布，以弥补传统基于计量模型的租金分析中样本量不足、要素量化方式主观、模型形式僵化等问题。研究方法：研究利用网络爬虫技术获取相关时空大数据，随后采用密度制图、空间句法分析等量化相关指标，并通过LASSO模型筛选影响因子，最后采用机器学习方法对社区尺度商铺租金进行空间布局分析。研究结果：基于机器学习方法的广州市中心城区社区尺度商铺租金布局分析能较好拟合估计值和观测值，其分析结果显示广州市中心城区中荔湾区、越秀区、天河区和海珠区各自形成了成熟的高租金核心，商铺租金从高值区域向外逐渐下降。研究结论：该研究方法可以应用于房租、房价空间分布估计等领域，研究成果可作为基准地价更新、城市异值空间发现等的参考。

A Machine Learning Approach for Community-scale Commercial Rents Mapping

The purpose of this paper is to apply a new set of methodology to obtain fine-scale commercial rent mapping for the central part of Guangzhou to overcome the obstacles in the traditional research on mapping commercial rents, including high cost of data acquisition, subjectivity in the quantification of variables and rigid model forms. The research methods are as follows. First, this research obtains spatial and rent-related data through web crawler technology. Second, influencing factors are quantified with kernel density estimation and space syntax analysis, and are filtered with LASSO model. The machine learning algorithms are used on the processed data set and their performance are compared before this research evaluates the commercial rents distribution in the study area with selected machine learning algorithm-RFR. The results show that each of the four districts of Guangzhou City, including Liwan, Yuexiu, Tianhe and Haizhu, has its own high-rent commercial centre and commercial rents gradually decline with the increase of the distance from the commercial centres. In conclusion, the proposed method can be used in the fields of the evaluation on housing rents and spatial distribution of housing prices. Moreover, the result of commercial rent mapping in this study can be used to renew the benchmark prices of land leasing and to discover the urban heterogeneous areas.