Footprint Generation using Fuzzy-Neighborhood Clustering

Geometric footprints, which delineate the region occupied by a spatial point pattern, serve a variety of functions in GIScience. This research explores the use of two density-based clustering algorithms for footprint generation. First, the Density-Based Spatial Clustering with Noise (DBSCAN) algorithm is used to classify points as core points, non-core points, or statistical noise; then a footprint is created from the core and non-core points in each cluster using convex hulls. Second, a Fuzzy-Neighborhood (FN)-DBSCAN algorithm, which incorporates fuzzy set theory, is used to assign points to clusters based on membership values. Then, two methods are presented for delineating footprints with FN-DBSCAN: (1) hull-based techniques and (2) contouring methods based on interpolated membership values. The latter approach offers increased flexibility for footprint generation, as it provides a continuous surface of membership values from which precise contours can be delineated. Then, a heuristic parameter selection method is described for FN-DBSCAN, and the approach is demonstrated in the context of wildlife home range estimation, where the goal is to a generate footprint of an animal’s movements from tracking data. Additionally, FN-DBSCAN is applied to produce crime footprints for a county in Florida. The results are used to guide a discussion of the relative merits of the new techniques. In summary, the fuzzy clustering approach offers a novel method of footprint generation that can be applied to characterize a variety of point patterns in GIScience