Poi Data Versus Land Use Data, Which Are Most Effective In Modelling Theft Crimes?
Jiajia Feng Department of Architecture, Tsinghua University
Yuebing Liang Department of Urban Planning and Design, The University of Hong Kong
Qi Hao Department of Architecture, Tsinghua University
Ke Xu Department of Architecture, Tsinghua University
Waishan Qiu Department of City and Regional Planning, Cornell University
Alleviating crime and improving urban safety is important for sustainable development of society. Prior studies have used either land use data or point-of-interests (POI) data to represent urban functions and investigate their associations with urban crime. However, inconsistent and even contrary results were yielded between land use and POI data. There is no agreement on which is more effective. To fill this gap, we systematically compare land use and POI data regarding their strength as well as the divergence and coherence in profiling urban functions for crime studies. Three categories of urban function features, namely the density, fraction, and diversity, are extracted from POI and land use data, respectively. Their global and local strength are compared using ordinary least square (OLS) regression and geographically weighted regression (GWR), with a case study of Beijing, China. The OLS results indicate that POI data generally outperforms land use data. The GWR models reveal that POI Density is superior to other indicators, especially in areas with concentrated commercial or public service facilities. Additionally, Land Use Fraction performs better for large-scale functional areas like green space and transportation hubs. This study provides important reference for city planners in selecting urban function indicators and modelling crimes.
Keywords: Poi; Land Use; Urban Functions; Theft Crime; Predictive Power; Sdg16