open access

https://hdl.handle.net/2115/47121

Type 2 Fuzzy Clustering Algorithm for Fuzzy Data

英語

Extension principle

2000

Proceedings of The International Conference on Measurement and Multivariate Analysis

Volume 1

15-17

International Conference on Measurement and Multivariate Analysis (ICMMA) and Dual Scaling Workshop. May 11, 12, 13 & 14, 2000 at The Banff Centre for Conferences Banff, Alberta, Canada

In many situations, precise measurements are vary costly, unnecessary, and even impossible to obtain in some cases. Thus, it often occurs that fuzzy data such as real intervals or real fuzzy numbers are available. In recent literatures, some method to handle fuzzy data have been developed [2, 4]. These are based on a parameterization representing centers and radiuses of fuzzy data. It is pointed out that different parameterizations of fuzzy sets can lead to different computational result even in very simple cases. In this paper, we propose other method to classify fuzzy data. This is based on a interval computations of α-cut sets of fuzzy data and a extension principle of fuzzy sets. This method enables us to obtain both type 2 fuzzy prototypes and type 2 fuzzy partitions. It requires high computing cost to determine exact intervals for a lot of α-cut sets. Therefore, we also propose a method to obtain approximate intervals which is based on a sensitivity analysis.

会議発表論文

Accepted Manuscript

BA49130456