Evolutionary optimization of information granules

Granulation involves decomposition of whole object into a collection of parts called granules. The granules are formed, based on the notions of indistinguishability, similarity, proximity or functionality. Building information granules, especially for highly dimensional data is a demanding task. In this study, we propose a genetic-based development of information granules. The approach is concerned with structural and parametric aspects of the information granulation that involves the number of information granules and their parameters. It is shown how information granulation supports a descriptive data analysis, namely a comprehensive process of revealing essential structures in data sets.