Noise Clustering via Dynamic Data Assigning Assessment

A new clustering algorithm that identifies clusters step by step is introduced. It is based on the principles of noise clustering dividing the data set into a good cluster and the remaining data that might contain only noise or also other clusters. The algorithm can be applied to finding just a few substructures (clusters), but also as an iterative method to data partition including the identification of the number of clusters and noise data. The algorithm is applicable in terms of both hard and fuzzy clustering techniques. An extended variant of the algorithm is developed in order to solve the problem of determining clusters that are sub-clusters of a certain separated cluster. The algorithm is applied to a gene expression data set and finds groups of coregulated genes.