An IMPROVED FCM-POSSIBLE CLUSTERING ALGORITHM FOR INTERVAL DATA

In order to overcome the disadvantages of Fuzzy c-Means(FCM) and possible clustering algorithm in the practical application of interval data, an improved FCM-possible clustering algorithm for interval data is proposed in this paper by combing their merits. The improved clustering algorithm introduce the possibility theory into the clustering problem of interval data, by relaxing the constraints of the sample membership and modifying IFCM algorithm's objective function The results of simulation experiments and the average CR index analysis show that: For the cluster problems containing poorly representative sample data such as noise and outliers, the improved FCM-possible clustering algorithm proposed in this paper is much better than the FCM algorithm and possible clustering algorithm, which can effectively reduce the influence on the clustering result by the noise