An Hybrid Algorithm for Microarray Gene Classification


  • Sharath Kumar Y. H., Narayan Naik


The high dimensionality of features with a comparatively small sample size was al- ways a challenging issue in microarray gene expression analysis. So, it is necessary to develop a robust and efficient feature selection technique to perform precise classification task on microarray datasets. A hybrid feature selection known as mRMRAGA technique was proposed, which combines the minimum redundancy and maximum relevance (mRMR) along with the Adaptive Genetic Algorithm (AGA). The minimum- redundancy-maximum-relevance (mRMR) technique has been often employed for identifying the characteristics of the gene and also their phenotype more accurately. The mRMR is the process in which feature relevance are narrowed down that has been de- scribed when pairing with their relevant feature selection. The Genetic Algorithm (GA) was inspired by the process of natural selection, which works on the basis of heuristic search technique. In the following section, an adaptive genetic algorithm has been used thatis the improvised versionof the Genetic Algorithm.

In this work, four benchmarked microarray gene expression datasets were used to conduct the experiment. These datasets contain a heterogeneous number of class labels, which is one dataset contains two class labels, whereas the remaining three contain more than two class labels.