Feature selection is one of the most important preprocessing steps for a data mining, pattern recognition or machine learning problem. Finding an optimal subset of features, among all the combinations is a NP-Complete problem. Lot of research has been done in feature selection. However, as the sizes of the datasets are increasing and optimality is a subjective notion, further research is needed to find better techniques. In this paper, a genetic algorithm based feature subset selection method has been proposed with a novel feature evaluation measure as the fitness function. The evaluation measure is different in three primary ways a) It considers the information content of the features apart from relevance with respect to the target b) The redundancy is considered only when it is over a threshold value c) There is lesser penalization towards cardinality of the subset. As the measure accepts value of few parameters, this is available for tuning as per the need of the particular problem domain. Experiments conducted over 21 well known publicly available datasets reveal superior performance. Hypothesis testing for the accuracy improvement is found to be statistically significant.

Feature selection is one of the most important preprocessing steps in data mining and knowledge Engineering. In this short review paper, apart from a brief taxonomy of current feature selection methods, we review feature selection methods that are being used in practice. Subsequently we produce a near comprehensive list of problems that have been solved using feature selection across technical and commercial domain. This can serve as a valuable tool to practitioners across industry and academia. We also present empirical results of filter based methods on various datasets. The empirical study covers task of classification, regression, text classification and clustering respectively. We also compare filter based ranking methods using rank correlation.