Wavelet‐Based Clustering Method for Geographical Flows Within a Linear Feature

作者：Meng Gao, Tao Pei, L. Jiang, Xiaorui Yan, Zidong Fang, Le Liu, Ya Fang · 发表于：Transactions in GIS · 年份：2025 · DOI：10.1111/tgis.70079 · 被引用次数：1 · 研究领域：Anomaly Detection Techniques and Applications、Time Series Analysis and Forecasting、Advanced Clustering Algorithms Research

ABSTRACT Geographical flows may indicate interactions between locations. If a set of flows has a higher flow volume and their origins and destinations are both concentrated, it can be considered a cluster. Identifying such clusters helps locate areas of intensive interactions, aiding the understanding of geographical patterns. However, unlike point clustering, a collection of flow data may demonstrate inhomogeneity regarding flow lengths. This refers to the frequencies of flows not being uniquely distributed across different length scales, where shorter flows often have a higher volume than longer ones. Nevertheless, mainstream flow clustering methods rarely consider such inhomogeneity, limiting further insights from clustering, as longer flows may be overlooked in the clustering process due to their lower volume. Here, we propose a wavelet‐based clustering method for flows within a linear feature. We used wavelet transforms to identify flow clusters in the spatial‐frequency domain. The taxi flows in Beijing's Chang'an Street were adopted as a case study to demonstrate the practicality of the approach. Some clusters with longer lengths were identified, indicating irreplaceable interactions between locations and having unique implications for transportation planning. Our study emphasizes the necessity of considering flow length in clustering and introduces a promising approach through spatial‐frequency domain analysis.