Fuzzy based Feature Selection Scheme through Transductive SVM Technique for Cancer Pattern Classification and Prediction

Abstract

Objectives: A reliable and precise classification of tumor types is of great importance and essential for successful Diagnosis and Drug Discovery. Methods: Gene expression profiling has shown a great prospective in the outcome prediction of different types of Cancers, with the innovation of Microarray Technology. Microarray cancer data, which have been organized as samples versus genes fashion, are being exploited for the tissue sample classification. They are also useful for identifying potential gene markers in each subtype of cancer, that helps to analyze a particular cancer type in a successful manner. Findings: In this paper a new method for classification based on Fuzzy Rough-Set Feature Selection Approach through Transductive SVM Technique, which is called as FFS+TSVM is proposed. Moreover recently, for Cancer Pattern Classification, a combined Consistency based Feature Selection Approach through Transductive Support Vector Machine (CBFS+TSVM) has proposed and its prediction accuracy has encouraged than that of existing other Classifiers. However, from the literature survey, it is revealed that the performance of the existing scheme can be improved if Fuzzy Rough Set Approach for Gene Feature Selection with Transductive Support Vector Machine is employed. The present work is implemented with the help of Bio Weka and studied thoroughly in terms of Computational Cost, Dimensionality Reduction, Threshold, Classification Error and Classification Accuracy. Applications/Improvements: The proposed work outperforms the existing Transductive Support Vector Machine (TSVM) in terms of Dimensionality Reduction, Threshold, Classification Error and Classification Accuracy for various Cancer Patterns.