In microarray gene expression analysis, a challenging issue has always been the feature's high dimensionality with a restricted sample size. For microarray datasets to be accurately classified, a reliable and effective feature selection method must be created. The maximum relevance (mRMR), minimum redundancy and adaptive Genetic Algorithm (AGA) are used in the hybrid feature selection technique known as mRMRAGA. The technique known as mRMR is widely used to more precisely determine the phenotypic traits of genes. The method by which feature relevance is reduced and described when paired with their pertinent feature selection is known as the maximum relative margin of rejection. Natural selection, which relies on heuristic search techniques, served as the model for the Genetic Algorithm (GA). The Adaptive genetic algorithms are genetic algorithms that have been modified and applied in the part that follows. In this paper, the experiment was carried out using four benchmarked microarray gene expression datasets. One of these datasets has two class labels, while the other three have more than two. This indicates that the number of class labels in these datasets is heterogeneous.