Abstract:
The growing self organizing map (GSOM), a variant of the self organizing map, is a dynamic feature map model used for knowledge discovery in high dimensional datasets. It has been used mainly to identify hidden patterns in static data in an unsupervised manner. Several extensions to the GSOM that enable dynamic data analysis have been proposed. In this paper we discuss such an extension and its capabilities in discovering time variant patterns in datasets of seismic activity. The results obtained by processing clusters generated by the GSOM using the data skeleton model and spread factor extensions, emphasize the usability of the GSOM in dynamic data analysis.
