An Extension of the K-Mean Algorithm for dealing with Dimensionality Curse

Multi-attribute clustering and dimensionality reduction is achieved through a generalization of the k-mean algorithm. Entropy analysis is utilized to reduce dimensionality and to find initial conditions or starting points for the extended k-mean algorithm. We start with the deciding criteria attribute, and based on comparison and selection of the surrounding points, inter value comparison and gain assessment is performed leading to optimal value selection. Dwarf data cubes are utilized to mitigate the effects of the curse of dimensionality, specifically in a database environment where suffixes and prefixes are present. Finally the use of the entropy generated result for the initial conditioning of the extended k-mean algorithm is explained