A review of feature selection algorithms that treat the microarray data redundancy

Abstract

In recent times, the redundancy analysis in attribute selection algorithms in machine learning has become a constant. Studies have shown that the percentages of prediction, after removing these attributes, are better than the cases where it is not. Furthermore, by excluding it from data set, the temporal complexity of the classifier is reduced because it has less data to process. In the actually, the algorithms have evolved in this regard and treat redundancy in different ways and with different criteria. The main aim of this review is to present the different evaluation criteria to address data redundancy in ADN microarrays. The study applied analysis-synthesis, historic-logical and inductivedeductive methods. We conducted a literature review of articles published since the 90's which contain algorithms to select attributes and take into account the dependency between them. The article describe a general way, his steps, the criterion used in the analysis of redundancy and some of its advantages and disadvantages.

Keywords

• Analysis of redundancy
• evaluation criteria
• feature selection
• filter algorithms.