A Clustering Based Feature Subset Selection Algorithm for High-Dimensional Microarray Data Using Fuzzy Entropy with Neuro-Fuzzy Classifier

Feature selection involves the process of selecting a subset of relevant features that produces the result as the original set of features. The central assumption of using a feature selection technique in high dimensional data is that the data may contain many redundant or irrelevant features. Microarray dataset may also contain a huge number of redundant (insignificant) and irrelevant features which lead to loss of information. One of the key processes is that attribute reduction; when the data set is used for classification without attribute reduction it produces the wrong results. The features that are presented in the data set are inconsistent and redundant. To progress the efficiency of classification these inconsistent and redundant features must be eliminated. Based on these criteria, a Fuzzy Entropy feature selection algorithm is proposed and evaluated. The features are extracted by the set of input patterns using membership function. The performance of Fuzzy Entropy feature selection algorithm is compared with the existing methods namely Sequential Feature selection and FAST algorithm. Along with this NeuroFuzzy classification algorithm is proposed to present comparative analysis to obtain accuracy. The accuracy of the proposed algorithm is compared with KNN and Fuzzy classifier.