The Performance Review of mRMR for Gene Selection and Classification of DNA Microarrays

Abstract

There are two main stages in the classification of DNA microarray data. The first stage is known as gene selection and the second stage is the classification of selected genes. The number of genes produced in high-dimensional microarrays is enormous, and only some of these genes help to identify a particular disease. The selection of relevant or informative genes that provide sufficient information about the condition is therefore essential. Gene selection is vital in reducing the data dimensionality which can ease the workload of the computer and increase the high classification performance. In this paper, we review the recent performance on one of the most popular filter based gene selection technique, maximum relevance minimum redundancy (mRMR). We also discuss several current improvements on the mRMR method.