An Adaptive Harmony Search Approach for Gene Selection and Classification of High Dimensional Medical Data

In Bioinformatics, microarray data analysis has gained substantial attention for disease diagnosis. Microarray data is represented with a huge search space which imposes the foremost difficulties in selection of most relevant facts in terms of genes. In this esteem, we have recommended a hybridised harmony search and Pareto optimization approach for feature selection in high dimensional data classification problem. In the first stage an adaptive harmony search algorithm for gene selection with probability distribution factor for optimal gene ranking is implemented. This selection is further refined applying a bi-objective Pareto based feature selection technique to select optimal minimum number of top ranked genes. The importance and relevance of the selected genes are verified through a few numbers of classifiers. Experimental analysis is conducted over four well known microarray datasets. Finally statistical analysis is conducted to prove the superiority of proposed work with two other nature inspired algorithms. Simulation result reveals that the proposed hybridisation is providing high potentiality in both sample classification and feature subset prediction prospective for high dimensional databases.