A global polycenter identification method with single‐source data: The integration of local multisource data recognition

Abstract

With the widespread application of multisource data, the identification of urban polycenters faces the challenge of increasing data costs. This study developed a cost‐effective model for identifying urban polycenters by employing a combination of the Random Forest algorithm and Local Moran's I index. Using point‐of‐interest data from Amap, our model was benchmarked against a multisource data model to verify its effectiveness and accuracy. The results indicate that the single‐source model possesses an accuracy comparable to that of the multisource model in determining the centrality and spatial distribution of urban centers, thus offering a substantial capability to reduce reliance on multisource data. The random forest method exhibits a significant accuracy advantage over traditional ordinary least squares regression methods. However, it also exhibited susceptibility to overfitting and variations in data sampling. This suggests that while the model is highly effective for large‐scale urban studies, it requires careful handling of data inputs. This model can be applied to actual urban planning and research, providing a useful instrument for investigating urban polycentric structures at different spatial scales. This will increase the usefulness of the model in real‐world scenarios and lower the expenses related to analyzing urban data.