V. CONCLUSION
This article presents a novel semi-supervised fuzzy K- NN classification algorithm for cancer classification form microarray gene expression data. The performance of the proposed semi-supervised method is compared with its two other supervised counterparts namely, K-NN and fuzzy K- NN classifiers and two non-fuzzy, non-NN based methods namely SVM and Naive Bayes classifier. The effectiveness of the proposed technique is tested using five microarray gene expression datasets for five different types of cancers. From the experimental observation it is found that the proposed semi- supervised method produces better accuracy in comparison to its other supervised counterparts when number of labeled patterns present in the microarray gene expression data is very less. This is due to the inclusion of the unlabeled samples (along with the limited training samples) during the training/learning phase of the proposed method as opposed to the traditional supervised methods. Also the use of the fuzzy system in the proposed semi-supervised fuzzy K-NN model helps to capture the overlapping classes as normally present in microarray gene expression data. In future, the proposed method may be tested extensively on other microarray gene expression cancer datasets. En- couraging results obtained from the proposed semi-supervised method may lead us to develop semi-supervised versions of the other existing classifiers in future.