The microarray technology permits simultaneous monitoring of several thousand gene expressions for each sample. However, classifying such samples into their distinct classes may be quite challenging owing to the higher number of genes (the features) over the actual samples. As a consequence, there is a need to make an investigation of the new and the robust techniques of machine learning that can accurately classify the microarray data. The B cells further give some humoral immunity by means of distinguishing into the Plasma Cells (PCs) that are antibody-secreting which is a process that needs a cellular division linked to the DeoxyRibonucleic Acid (DNA) hypomethylation. Classification of cancer is another critical issue in the analysis of cancer treatment. A very effective method in the classification of cancer is its gene selection. But choosing the gene subset that increases the accuracy may be a Non-Deterministic Polynomial (NP)-hard problem. For this work, a methodology was proposed for choosing an ideal set of genes which permit classification of the class of disease based on gene expression along with a good accuracy. The proposed method was compared with feature selection method that employed the Genetic Algorithm (GA).The feature selection of the proposed work is using the Artificial Bee Colony (ABC) algorithm with feature selection that is Levy search based. The results have proved that this new technique had an increase in accuracy of classification which was compared to the Genetic Algorithm and also the Artificial Bee Colony based technique. This also had a very fast convergence. © 2020, World Academy of Research in Science and Engineering. All rights reserved.