Multi-Objective Firefly Algorithm for Multi-Class Gene Selection

Abstract

Multiclass cancer classification is an emerging technique which presents the possibility of cancer identification using microarray data. For selecting genes in the multiclass gene categorization filter methods are frequently used. But the filter method is not applicable for some of the multiclass microarray data sets because of the rigorous heterogeneity of biological tissues and samples. So, for selecting genes decay the multiclass ranking statistics into class explicit statistics and then Pareto-front analysis is used. Also, to identify the Pareto-optimal set the non-dominant sorting genetic algorithm is suggested. But the drawback is this method does not scale with high complexity. Because, where the number of elements which are represented to mutation is large there is an exponential raise in search space size. So, in this manuscript an innovative technique is introduced which is called Multiobjective Firefly Algorithm for Multiclass Gene Selection (MFGS). A firefly has a tendency to be fascinated towards other fireflies with superior flash intensity. The multiple objective firefly algorithms intend to optimize two or more conflicting characteristics represented by fitness functions. In the multiple objective firefly method, a set of Pareto-optimal solutions are created which concurrently optimize the contradictory necessities of the multiple fitness functions. In the proposed method, the genes are selected by optimizing the number of fireflies in the multiple class-specific statistics. An experimental result shows that when compared to the existing method, there is less complexity, high classification accuracy of the proposed MFGS method.