Metaheuristic-Based Hybrid Feature Selection Models

1. Introduction

Microarray technology produces high dimensional datasets by measuring the expression levels of tens of thousands of genes in a single experiment under varying conditions. It has become an indispensable tool for biological, medical and pharmaceutical researchers to get a better understanding of the diseases at genomic level. On the other hand, the inherent problem i.e., large number of features and small sample size of microarray dataset makes the analysis process difficult for the problem. Typically, are latively small numbers of features are found to be strongly correlated with the phenotypes in question? Therefore, to identify these discriminative features from gene expression dataset, a data mining tool (Witten, Frank & Hall, 2011) known as feature selection technique plays an important role. The methods specifically used for feature selection can be categorized into two major groups namely, filter and wrapper methods (Saeys, Inza&Larranaga, 2007; Guyon, Nikravesh, Zadeh, 2006; Dash& Patra, 2016a).

Filter methods select features considering individual characteristic of each feature without taking into account the mutual dependencies among features. Then the features are sorted by their assigned ranks. The top ranked features are kept for further analysis by removing low ranked features. Actually, these selected features are used to develop the diagnostic model to efficiently predict the diseases. On the contrary, in wrapper methods, a search algorithm is wrapped around a learning algorithm: so that an estimated learning accuracy for all subsets can be calculated to derive an optimal one. This method is computationally intensive in comparison to filter methods because to obtain an optimal subset of features all possible subsets need to be examined which is practically a difficult task. Thus, to alleviate this difficulty in wrapper methods a metaheuristic search strategy (Ghosh & Jain (Eds),2005; Akadi, Amine, Ouardighi & Aboutajdine, 2009; Dash, 2016) may be adopted from a set of heuristic or stochastic algorithms which can be able to generate an optimal subset effectively.

In the gene selection process, an optimal gene subset is always relative to a certain criterion. Several information measures such as entropy, mutual information (Ding& Peng, 2005)and f-information (Maji, 2009) have successfully been used in selecting a set of relevant and non redundant genes from a microarray data set. An efficient and effective reduction method is necessary to cope with large amount of data by most techniques. Growing interest in developing methodologies that are capable of dealing with imprecision and uncertainty is apparent from the large scale research that are currently being done in the areas related to fuzzy (Zadeh, 1965) and rough sets. Rough set theory (RST) was introduced by Pawlak (1982) and has been used widely by researchers as a classifier and selection technique. The success of rough set theory is due to three aspects of the theory. First, only hidden facts in the data are analyzed. Second, additional information about the data is not required for data analysis. Third, it finds a minimal knowledge representation for data. Due to this Rough set theory is used in complement with other concepts such as, fuzzy set theory. The two fields may be considered similar in the sense that both can tolerate inconsistency and uncertainty. The only difference among these two fields is the type of uncertainty and their approach to it. While fuzzy sets are concerned with vagueness, rough sets are concerned with indiscernibility. The fuzzy- rough set-based approach considers the extent to which fuzzified values are similar.