兴趣点数据的城市功能区识别方法对比  

Comparison of identification methods for urban functional areas based on point of interest data

作　　者：崔方迪 袁璞 CUI Fangdi;YUAN Pu(th Research Institute,China Electronics Technology Group Corporation,Beijing 100081,China)

机构地区：[1]中国电子科技集团公司第十五研究所,北京100081

出　　处：《北京测绘》2024年第12期1652-1658,共7页Beijing Surveying and Mapping

基　　金：北京市自然科学基金(8222011)。

摘　　要：城市功能区作为社会经济活动载体,对城市资源配置及规划管理具有重要意义。传统城市功能区识别方法存在主观性强、效率低、识别精度低等缺陷。鉴于此,本研究引入自然语言处理(NLP)技术中的潜在狄利克雷分布(LDA)模型和词频-逆文档频率(TF-IDF)模型,挖掘城市兴趣点(POI)数据语义信息,揭示区域性潜在的功能利用模式。首先,城市空间被分割为500 m×500 m粒度的格网,将POI数据映射到对应的地理网格单元,并基于词袋模型构建语料库。随后,分别采用LDA模型和TF-IDF模型计算格网单元和POI数据之间的分布模式来识别城市功能区。最后,将城市功能区识别结果与百度电子地图及街景影像进行比对来评估精度。实验结果表明,LDA模型算法精度达78%,优于精度为63%的TF-IDF模型。LDA算法能更加准确地识别城市功能区的功能利用类型,且能够在TF-IDF模型较难区分的功能区上取得较好识别效果。本研究揭示了POI数据与城市功能区之间的潜在语义关系,可作为城市功能区研究的参考和补充,可辅助城市规划者动态监测城市结构,对未来城市更新与发展进行布局引导。As the carrier of social and economic activities,urban functional areas are of great significance to urban resource allocation and planning management.The traditional identification method for urban functional areas has the defects of strong subjectivity,low efficiency,and poor accuracy.In view of this,this paper introduced the latent Dirichlet allocation(LDA)model in natural language processing(NLP)and the term frequency-inverse document frequency(TF-IDF)model to explore semantic information of urban point of interest(POI)data and reveal regional potential function utilization patterns.Firstly,the urban space was divided into a 500 m×500 m granular grid,and the POI data was mapped to the corresponding geographical grid units.The corpus was built based on the bag of word model.Secondly,the LDA model and TF-IDF model were used to calculate the distribution patterns between grid units and POI data,so as to identify urban functional areas.Finally,the identification results of urban functional areas were compared with the Baidu electronic map and street view images to evaluate the accuracy.The experimental results show that the accuracy of the LDA model is 78%,which is higher than 63%of the TF-IDF model.The LDA algorithm can identify the function utilization patterns of urban functional areas more accurately and can achieve better identification effects in the functional areas that are hard to distinguish by the TF-IDF model.This paper reveals the potential semantic relationship between POI data and urban functional areas,which can be used as a reference and supplement for the research on urban functional areas.It can also assist urban planners in dynamically monitoring urban structure and guide the layout of future urban renewal and development.

关 键 词：城市功能区 兴趣点(POI) 自然语言处理(NLP) 词频-逆文档频率模型(TF-IDF) 潜在狄利克雷分布(LDA) 

分 类 号：P208[天文地球—地图制图学与地理信息工程]