On semi-supervised fuzzy c-means clustering
Yasunori Endo, Yukihiro Hamasuna, Makito Yamashiro, Sadaaki Miyamoto · 2009
We have two methods of pattern classification, one is supervised and the other is unsupervised. Unsupervised classification, which is called clustering and classifies data except external criteria, is very useful in the methods of pattern classification so that it has been applied in many fields. There are two types of clustering, one is hierarchical and the other is non-hierarchical. We often use hard c-means clustering (HCM) or fuzzy c-means clustering (FCM) as typical methods of non-hierarchical clustering. By the way, supervised classification can achieve practical classification results but can't handle a lot of data. On the other hand, unsupervised classification can handle a lot of data but the method is complex and sometimes results look a bit of strange. Therefore recently, study of semi-supervised classification has been studied. This classification has advantages of both of the above-mentioned methods, e.g., practical results, low costs and short calculation time. In this paper, we propose new semi-supervised classification algorithms based on fuzzy c-means clustering in which some membership grades are given as supervised membership grade in advance.