Extended Gustafson–Kessel granular hotspot detection

Abstract

A hierarchical granular model that includes a variation of the Extended Gustafson–Kessel (EGK) clustering algorithm for massive event data sets applied in hotspot analysis is proposed. We construct a granular view of the distribution of hotspots on a geographic map related to a phenomenon, whose data are collected in a massive data set. To obtain the final information granules, we partition randomly the data set in s chunks and execute the EGK clustering algorithm separately to each chunk in a distributed architecture. Finally, a weighted EGK clustering algorithm is applied to a data set formed by the centers of all the local hotspots to the elliptical hotspots which constitute the upper level information granules giving a global overview of the spatial distribution of the hotspots on the map. Two indices are calculated to assess how justifiable is this granular view in terms of spatial distribution of the final hotspots on the map. A set of tests on the massive data set Archive Fire from Indonesia are performed by setting a threshold for the above two indexes and by varying the number of chunks. The results show that the proposed algorithm provides a justifiable granular view of the hotspot detected on the map. Comparison tests with respect to other clustering hotspot detection algorithms show that the proposed method is very efficient in terms of execution time and spatial distribution of the final hotspots.