High Dimensional Object Analysis Using Rough-Set Theory and Grey Relational Clustering Algorithm

Abstract: High dimensional feature selection and data assignment is an important feature for high dimensional object analysis. In this work, we propose a new hybrid approach of combining attribute reduction of the Rough-set theory with Grey relation clustering. Designing clustering becomes increasingly tougher task as the dimensionality of the data set increases. Previously constraint based clustering algorithms that satisfy user specified constraints have been used for high dimensional data sets. Such algorithms suffer from serious limitations and can introduce biases of the user, thus obscuring discovery of clusters and hidden relations in the data set. In this work, we transform the high relevance values into the same class using Grey relation to give an appropriate cluster of information, which we process through Rough set to reduce attributes. We use this approach to analyze the data of plant diversity from North America and find that ground elevation and species numbers can capture the most important attributes of the data set. This analysis of ecological data presents a proof of principal for the novel hybrid approach using Grey relational clustering and Rough set theory.

Keywords: RST, GRA, Rule Generation, High Dimensionality, Indiscernibility (IND).