Semi-supervised Kernel-Based Fuzzy C-Means

This paper presents a semi-supervised kernel-based fuzzy c-means algorithm called S 2 KFCM by introducing semi-supervised learning technique and the kernel method simultaneously into conventional fuzzy clustering algo- rithm. Through using labeled and unlabeled data together, S 2 KFCM can be ap- plied to both clustering and classification tasks. However, only the latter is con- cerned in this paper. Experimental results show that S 2 KFCM can improve classification accuracy significantly, compared with conventional classifiers trained with a small number of labeled data only. Also, it outperforms a similar approach S 2 FCM.