Feature Selection for Gene Expression Data Analysis – A Review

Gene selection in microarray data analysis is defined as the process of identifying a small number of informative and relevant genes that can find any sample from the dataset into the correct class. The feature selection process is categorized into three types: wrapper, embedded and filter techniques. Filter methods use statistical ranking for feature selection by ordering the features individually. They select the relevant features independent of any supervised learning algorithm. The wrapper techniques use a number of search methods to evaluate the possible subset of important features. From that it selects the subset of features that gives the best classification accuracy. In embedded methods, feature selection methods are incorporated in the training process. This paper reviews several feature selection methods used to find significant features from gene expression data for use in classification.