MRMR-GWICA: A hybrid gene selection and ensemble clustering framework for breast cancer gene expression data

Microarray or gene expression profiling is conducted in a single experiment with different types of cells or tissue samples to evaluate and compare the extent of gene expression. The role of classifying the sample in question is more promising as the microarray dataset is disrupted due to the dimensional component, with a limited number of samples together with incorrect and noise genes. This is most important for the screening and diagnosis of a sample of breast cancer. The method of gene selection is currently being formulated to classify a few numbers of significant genes that are correlated with the extremely predictive process in the classification field. To carry out this method, a revolutionary genetic selection algorithm, namely the minimal Redundancy Maximal Significance (mRMR), is devised and integrated with the Gene Weight Imperialist Competitive Algorithm (GWICA), mRMR-GWICA, to classify insightful genes from the microarray-containing profile. The ground-breaking approach relies on the parallel Progressive Inductive Subspace Ensemble Clustering (PPISEC) algorithm to measure the precision of the classification of the known genes. PPISEC-WICA algorithm has three key steps: Improved Support Vector Machine (ISVM) classifier and Incremental Ensemble Member Chosen (IEMC) by GWICA are used to pick the centroid values and to execute the gene expression data clustering a structured cut algorithm is suggested. Experimental findings reveal that the formulated PPISEC system performs well on data containing the expression of the gene for breast cancer relative to conventional clustering ensemble approaches. This design is applied in a dataset comprising the gene expression dataset (GSE45827) which is collected from breast-cancer-gene-expression-cumida. The outputs of the clustering method are determined based on certain metrics such as precision, recall, f-measurement, accuracy, Adjusted Rand Index (ARI), and Normalized Mutual Information (NMI). The clustering design is implemented by implementing the MATLAB simulation environment in the R2014a edition.