A Survey on Feature Selection and Extraction Techniques for High-Dimensional Microarray Datasets

Abstract

In recent years, lots of data are generated and stored in the field of information technology, bioinformatics, text mining, face recognition, microarray data analysis, image processing, etc. From this microarray gene expression, data analysis gained the more importance due to role of disease diagnosis and prognoses to choose the appropriate treatment to the patients. Generally, gene expression data are a sort of high-dimensional data with small number of observation and large number of attributes. Interpreting the results from the gene expression data are difficult one due to the “curse of dimensionality.” For this issue, dimensionality reduction plays an important role, since it reduces the number of variables by using the techniques such as feature selection and feature extraction. The main aim of these approaches is to reduce/downscale the high-dimensional feature space to low-dimensional representation with an affection of classification accuracy. For this concern, the objective of this chapter is to gather and provide the up-to-date knowledge in the field of feature selection methods applied in the microarray data analysis as possible for the readers. In this chapter, a brief introduction about feature selection methods in the DNA microarray analysis was presented in Sect. 1. The taxonomy of the dimensionality reduction methods is represented in the diagrammatic way. Five feature selection methods such as filter, wrapper, embedded, hybrid, and ensemble methods have been discussed in detailed manner and tabulated the recent proposed algorithms, datasets used and accuracy achieved in the respective methods, and also the advantages and disadvantages of each method are discussed in Sect. 2. In Sect. 3, supervised, unsupervised, semi-supervised gene selection methods, with advantages and disadvantages were discussed. Section 4 discusses the feature extraction techniques applied in the microarray data analysis. Finally, Sect. 5 provides intrinsic characteristics of microarray data with respect feature selection.