统计与数据科学学院知识库

  进入“十四五”规划时期，面对我国城市发展不均衡、不充分的现状，如何优化公共服务设施布局，助力解决相关问题成为当前城市规划研究的热点。近年来得益于智能移动设备的普及，海量的人类活动数据得以保存，形成结构多样的多源大数据。通过机器学习与空间统计分析，挖掘多源大数据中的信息，并以此为基础进行公共服务设施布局的优化设计是一种有效手段。将城市遥感影像数据、POI（Point of interest）数据、人口统计数据整合为多源数据集，从城市空间特征识别的角度出发，对城市功能区、设施服务可达性测度、设施选址推荐三个角度进行分析研究，旨在对城市公共服务设施的空间布局进行更为精准的识别和综合优化。主要工作包括以下三个方面：

  首先，基于Mask R-CNN算法和样方密度法，以兰州市主城区为例，通过挖掘遥感影像数据的自然特征和POI数据中的人文特征，识别建筑物轮廓和相关功能信息，对城市功能区进行划分并识别其功能类型。相较于单一视角和数据进行的功能区识别，结果更加精准、客观，更加贴合城市现状。

  其次，基于改进的两步移动搜索法，对兰州市主城区医疗服务设施的空间可达性进行测度。利用第七次人口普查数据、POI数据、统计年鉴数据等，划分三级医疗服务设施，分级构建服务半径，从供需角度出发测算服务距离，旨在更加精准的反映当前研究区内医疗设施的空间可达性情况，也更加贴合医疗设施的服务方式。结果显示当前研究区内的医疗设施空间可达性总体较好，但仍然存在部分区域设施空间分布不均衡的情况。

  最后，利用随机森林算法的分类功能，分析兰州市主城区医疗服务设施的分布情况并进行选址推荐。通过引入机器学习算法，挖掘POI数据中医疗设施的空间分布特征，并以此为基础预测不同区域医疗设施分布的推荐度，降低了选址推荐模型的主观性。结果显示选址推荐度较高的区域大多分布在城市郊区，与实际情况基本相符。对兰州市主城区医疗设施空间分布特征进行综合考虑后，设计兰州市主城区医疗设施布局优化方案。

    As we enter the "14th Five-Year Plan" period, optimizing the layout of public service facilities to help address the issue of unbalanced and inadequate urban development in China has become a hot topic in urban planning research. In recent years, the widespread use of intelligent mobile devices has enabled the storage of massive amounts of human activity data, resulting in structurally diverse multi-source big data. By using machine learning and spatial statistical analysis to extract information from multi-source big data, optimizing the design of public service facility layout becomes an effective means. This thesis combines multiple data sources, including urban remote sensing image data, Point of Interest (POI) data, and demographic data, to analyze and study urban functional areas, facility service accessibility, and facility location recommendations from the perspective of urban spatial feature recognition, achieving more precise recognition and comprehensive optimization of the spatial layout of urban public service facilities. The main work includes the following three aspects:

    Firstly, based on the Mask R-CNN algorithm and sample density method, taking Lanzhou's main urban area as an example, this article mines the natural features of remote sensing image data and the humanistic features of POI data to identify building outlines and relevant functional information, dividing the urban functional areas and identifying their functional types. Compared with the identification of functional areas using a single perspective and data, this method achieves more accurate and objective results, better fitting the urban reality.

    Secondly, using an improved two-step Floating Catchment method, this article measures the spatial accessibility of medical service facilities in Lanzhou's main urban area. By using data from the seventh national census, POI data, and statistical yearbooks, three levels of medical service facilities are defined, and service radii are constructed to calculate service distances from the supply and demand perspectives. This method aims to more accurately reflect the current spatial accessibility of medical facilities in the research area, and it better fits the service mode of medical facilities. The results show that the overall spatial accessibility of medical facilities in the research area is good, but there are still some areas where the distribution of facilities is uneven.

    Finally, by using the classification function of the random forest algorithm, this thesis analyzes the distribution of medical service facilities in Lanzhou's main urban area and makes location recommendations. By introducing machine learning algorithms and extracting the spatial distribution characteristics of medical facilities from POI data, this method predicts the recommendation degree of medical facility distribution in different regions, reducing the subjectivity of the location recommendation model. The results show that the regions with high location recommendation degrees are mostly located in the suburbs of the city, which is consistent with the current situation. After considering the comprehensive spatial distribution characteristics of medical facilities in Lanzhou's main urban area, an optimized layout plan for medical facilities in Lanzhou's main urban area is proposed.