基于互联网数据的城市社区租金评估及空间格局制图——以深圳市为例

准确刻画精细化尺度下的城市房租空间格局，对于研究城市居住行为、提高城市规划合理性十分重要。文章提出了一种基于互联网房租数据作为可靠数据源的城市房租空间格局制图方法。以深圳市作为研究区，通过广泛采集开放平台中用户发布的租房信息，绘制社区行政区尺度的房租空间分布图。房租空间制图涉及到对于没有样本数据区域平均房租的评估，因此，选取一系列与房租相关的房产属性、房屋区位及配套设施的评价指标，采用前馈神经网络技术构建评估模型。以2015年深圳市的住宅租赁市场作为研究对象，通过对结果的分析，以及与权威部门发布的统计数据进行比较，表明此方法能够有效地绘制社区尺度下城市房租的空间分布，模型预测结果的误差（%RMSE）为13.87%。所使用的互联网房租数据、POIs数据及前馈神经网络的建模工具均是开源的，而且所提出的方法论具有普适性，能够应用于其他研究区的房租空间格局制图，具有实践意义。

Unit Rent Appraisal in Community-scale and Spatial Pattern Mapping in a Metropolitan Area Using Online Real Estate Data: A Case Study of Shenzhen

Accurate mapping of the fine-scale spatial pattern of urban housing rent is crucial to study the urban residential behavior and to improve the rationality of urban planning. This paper proposed the use of online rental data as a reliable new data source for mapping a housing rent spatial pattern. Utilizing individual rental information from one of the biggest online house trading and renting platforms, Anjuke, we attempted to map the urban housing rent of the community unit, which is the most precise administration unit in China. Due to the online rental data not being evenly distributed throughout the region, we conducted a neural network-based hedonic model to predict the housing rent of unknown community units with six evaluation indicators: environmental conditions, traffic systems, job opportunities, living facilities, education, and health care. The 2015 housing rental market of Shenzhen was chosen as a study case. The prediction error (% RMSE) of the hedonic model-which is trained with online rental data-is 13.87%. The mapping result is consistent with the real spatial pattern of housing rent, and the average rent across the whole study area is 60.56 yuan/m2, 5.69 yuan/m2 less than the official statistical average. The case study results indicated that using online rental data can effectively map the spatial pattern of housing rent on a community scale. There are three significant benefits of using this online data to build the model: First, the proposed online data-based rent mapping method can be applied universally, as all the data as well as the model used in the proposed mapping method are generally open-access. This means that other researchers can easily use this method to map the housing rent or prices of a different study area. Second, there is a much larger volume of online rental data than what can be obtained through survey samples or real rental samples. Furthermore, the data collection cost is much lower than that of traditional data sources. Third, the update cycle of online rental data is more frequent, meaning that the urban rental market can be monitored at a high spatial-temporal resolution if this online data-based method is used. The outcomes of utilizing this method can also supply fundamental data concerning urban social issues studies such as housing burden, urban poverty, and residential segmentation.