Space-Time Hierarchical Clustering for Identifying Clusters in Spatiotemporal Point Data

Abstract

Finding clusters of events is an important task in many spatial analyses. Both confirmatory and exploratory methods exist to accomplish this. Traditional statistical techniques are viewed as confirmatory, or observational, in that researchers are confirming an a priori hypothesis. These methods often fail when applied to newer types of data like moving object data and big data. Moving object data incorporates at least three parts: location, time, and attributes. This paper proposes an improved space-time clustering approach that relies on agglomerative hierarchical clustering to identify groupings in movement data. The approach, i.e., space–time hierarchical clustering, incorporates location, time, and attribute information to identify the groups across a nested structure reflective of a hierarchical interpretation of scale. Simulations are used to understand the effects of different parameters, and to compare against existing clustering methodologies. The approach successfully improves on traditional approaches by allowing flexibility to understand both the spatial and temporal components when applied to data. The method is applied to animal tracking data to identify clusters, or hotspots, of activity within the animal’s home range.

Identification of Co-Clusters with Coherent Trends in Geo-Referenced Time Series

“…(2) decreasing; (3) first increasing and then decreasing; (4) first decreasing and then increasing; (5) first increasing, then decreasing, and finally increasing; (6) first decreasing, then increasing, and finally decreasing; (7) first decreasing, then increasing, decreasing, finally increasing. Comparisons with co-clusters of similar values show that the two co-clustering results were different: spatial distributions of station-clusters in BCC_MSSR were more stretched in space, and BCC_MSSR discovers coherent spatio-temporal patterns in local regions and certain time periods.…”

“…Overall, there are mainly seven types of coherent temperature trends in the representative co-clusters, ranked by the degree of complexity, as follows: (1) increasing; (2) decreasing; (3) first increasing and then decreasing; (4) first decreasing and then increasing; (5) first increasing, then decreasing, and finally increasing; (6) first decreasing, then increasing, and finally decreasing; (7) first decreasing, then increasing, decreasing, andfinally increasing. As displayed in Figure 7, the fourth to eighth co-clusters all exhibit increasing temperature trends, even with various ranges of temperatures and differences in increasing speed.…”

“…On one hand, these massive amounts of data provide opportunities to investigate complex patterns and knowledge to help decision-making [3]; on the other hand, how to extract patterns from these data becomes a challenging issue [4]. As one of the most important data mining tasks, clustering methods group data elements into clusters by identifying similar ones and separating dissimilar ones, thus helping to extract underlying patterns in the data [5][6][7].…”

Spatial Distribution Characteristics of Heavy Metals in Surface Soil of Xilinguole Coal Mining Area Based on Semivariogram

“…In the future research, the temporal and spatial differentiation of heavy metals should be discussed further. As the spatiotemporal Kruger method is suitable for the analysis of location, time, and multiattribute motion data [37], it can realize the assimilation of multi-source spatiotemporal datasets [38], such as those from tracking research of animals, plants, and spatiotemporal phenomena in different historical periods [39,40].…”

Geo-Marketing Segmentation with Deep Learning

“…Summary of clustering classifications and the common algorithms used to achieve partitioning, hierarchical, density-based, or grid-based approaches[66].…”