Recurrent authenticity of the clustering of great tribute to the function of special type
Keyword(s):
Big Data
◽
A method of credibilistic fuzzy clustering is proposed for problems when data are fed sequentially, in online mode and forms large arrays (Big Data). The introduced procedures are essentially gradient algorithms for optimizing the objective function of a special type, and have a number of advantages over known probabilistic and possible approaches and, above all, robustness to anomalous observations. The approach is based on similarity measure, parameters of that are determined automatically in the process of self-learning. The proposed procedures are a generalization of the known methods, characterized by high speed and simple in numerical implementation.