A new SVM integrated rough type-II fuzzy clustering technique

Clustering algorithms based on type-I fuzzy set theory have been used for handling overlapping partitioning area over the last few decades. However, these fail to deal with additional degree of fuzziness within the real life datasets, because the membership values of type-I fuzzy set are crisp real numbers. Therefore, since inception, the type-II fuzzy set theory has been studied to address the weakness of type-I fuzzy set theory as the membership value itself is fuzzy in type-II fuzzy set theory. On the other hand, rough set based clustering method helps in great extend to handle the inherent uncertainty and vagueness of the data with the concept of lower and upper approximation. However, in rough clustering, rough points are not so certain to a particular cluster. In that case, machine learning technique such as Support Vector Machine can be used to assign the rough points into proper clusters in order to get the better clustering result. Hence, in this article, Support Vector Machine integrated Rough type-II Fuzzy C-Means clustering technique using both the rough set and type-II fuzzy set theories, is proposed. The effectiveness of the proposed clustering technique has been demonstrated quantitatively and visually on several synthetic and real life datasets in comparison with other well-known clustering techniques. Finally, the superiority of the results produced by the proposed technique has been shown using statistical significance test.