Spread Control for Huge Data Fuzzy Learning

The control of Growing Self Organizing Maps (GSOM) algorithms presents a serious issue with huge data learning. In conjunction with the growing threshold (GT), the spread factor (SF) is used as a controlling measure of the map size during the growing process. The effect of the spread factor in fuzzy learning with Fuzzy Multilevel Interior GSOMs (FMIG) algorithm is investigated. Further analysis is conducted on very large data in order to demonstrate the spread control of data distribution with FMIG learning in comparison with Multilevel Interior Growing SOM (MIGSOM), GSOM, Fuzzy Kohonen Clustering Network (FKCN) and fuzzy GSOM. Therefore, the aim of this paper is to study the effect of the spread factor values on the map structure in term of quantization error, topology preservation and dead units. Experimental studies with huge synthetic and real datasets are fulfilled at different spread factor values for the advertised algorithms.