Identifying urban functional zones by capturing multi-spatial distribution patterns of points of interest. Issue 1 (31st December 2022)

• Record Type: Journal Article
• Authors: Qin, Quan; Xu, Shishuo; Du, Mingyi; Li, Songnian
• Is Part Of: International journal of digital earth. Volume 15:Issue 1(2022)
• Page Start: 2468
• Page End: 2494
• Publication Date: 2022-12-31
• Subjects: Urban functional zone -- point of interest -- spatial distribution pattern -- natural language processing -- word2vec; Geographic information systems -- Periodicals; Sustainable development -- Information technology -- Periodicals; Social planning -- Information technology -- Periodicals; 910.285
• DOI: 10.1080/17538947.2022.2160841
• Languages: English
• ISSNs: 1753-8947

Abstract

ABSTRACT: Urban Functional Zone (UFZ) identification is vital for urban planning, renewal, and development. Point of Interest (POI), as one of the most popular data in UFZ studies, is transformed into a geo-corpus under specific sampling strategies, which can be used with Natural Language Processing (NLP) technology to extract geo-semantic features and identify UFZs. However, existing studies only capture a single spatial distribution pattern of POIs, while ignoring the other spatial distribution information. In this paper, we developed an integrated geo-corpus construction approach to capture multi-spatial distribution patterns of POIs that were represented by different modal POI embeddings. Subsequently, random forest model was leveraged to classify UFZs based on those embeddings. A set of combination experiments were designed for performance validation. The results show that our proposed method can effectively identify UFZs with an accuracy of 72.9%, with an improvement of 8.5% compared to the baseline methods. The outcome of this study will help urban planners to better understand UFZs through investigating the integrated spatial distribution patterns of POIs embedded in UFZs.