A New Clustering Algorithm Using Interval Type-II FCM

Clustering analysis is an important concept in pattern recognition exactly when there is no available information about class labels of a dataset. Many researchers have worked on clustering and have tried to propose and improve clustering algorithms. One of the most famous clustering algorithms is Fuzzy C-Means which uses Fuzzy logic to find reliable cluster centers in a pattern set. As it uses fuzzy logic, it is a method of clustering which allows a point of data to belong to two or more clusters with some membership degrees. Despite FCM is a powerful algorithm, it suffers from outliers and cannot perform well when there are uncertainties in data or when it faces the clusters with different volumes or densities. IT2FCM is another clustering algorithm to overcome this problem using Interval Type 2 Fuzzy Logic. In this paper, we have used density concept in IT2FCM and proposed a method that improves its accuracy. We have also tested our method using several artificial and real UCI datasets and we have computed the accuracy for each data set using confusion matrix. The results show that our method is a better method than IT2FCM and FCM methods for most of tested data sets.

کلیدواژه ها:

Datamining ، fuzzy ، fuzzy type II ، clustering