A new approach of microarray data dimension reduction for medical applications

Abstract

To employ and develop the performance of the dimensionality reduction for microarray data there is need of good dimension reduction technique. High-dimensional data bring great challenges in terms of computational complexity and classification performance. Therefore, it is necessary to effectively compress in a low-dimensional feature space from high dimensional feature space to design a learner with good performance. Feature extraction has a stronger ability to extract structure information in variables. Feature selection preserves the original features so that obtained feature subset has better explanatory ability. Therefore, dimension reduction is essential to study and understand the mechanism of practical problems of the microarray data. Dimension reduction is the important term which is majorly used in the big areas of genetics, medical and bioinformatics field. In medical applications for high dimensional cancer microarray data the dimension reduction is the important step. In this paper, a new Maximal Information-based Nonparametric Exploration method is proposed for the dimension reduction of the microarray data. In MINE method the MIC (Maximal Information Coefficient) plays the important role to show the relation between the data. The paper focused on improving the performance in terms of recognition accuracy, relevance, interpretability and redundancy, after comparing the performance of MINE method and Total PLS algorithm on data.