A novel fuzzy clustering urban hot spot detection method—an application in crime analysis

Urban hotspot detection methods represent a significant feature in many urban analysis problems because they allow to detect where a phenomenon has intensified and how it evolves over time. K-means and Fuzzy C-means clustering algorithms are frequently used in urban hotspot detection because they have the advantage of being computationally fast; however, they are not very robust to the presence of noise and outliers and fail to capture hotspots that are not well separable. In order to increase the accuracy of hotspot detection methods based on the Fuzzy C-means algorithm, a novel hotspot detection method is proposed that uses the weighted Fuzzy C-means algorithm to evaluate the impact of the data points density. The proposed algorithm has the advantage of improving the performance of Fuzzy C-means-based hotspot detection algorithms, increasing robustness to data noise and capturing overlapping and elongated hotspots, without reducing computational speed. Experimental tests were carried out on crime analysis events that occurred in the District of Columbia (USA) between 2019 and 2024. They highlighted the ability of the proposed hotspot detection method to capture intersecting and elongated hotspots and to provide a valid tool for localization and analysis of the temporal evolution of urban hotspots.