Optimized feature selection method using particle swarm intelligence with ensemble learning for cancer classification based on microarray datasets

Abstract

Cancer is considered a leading cause of mortality in both developed and developing countries. Cancer classification based on the microarray dataset has provided insight into possible treatment strategies. A complicated and high-dimensional number of genes and a few numbers of instances are characteristics of the microarray datasets. Gene selection is therefore a challenging and required task for the data analysis of microarray expression. The selection of genes may reveal insight into the underlying mechanism of a particular biological phenomenon. Several academics have recently developed methods of feature selection, utilizing metaheuristic algorithms for interpreting and analyzing microarray data. Nevertheless, due to the few numbers of samples in microarray data compared to the high dimensionality, several data mining approaches have been unsuccessful to select the most relevant and informatics genes. As a result, incorporating various classifiers can enhance feature selection and classification performance. The current study aims to propose a method for cancer classification by employing ensemble learning. Hence, in this paper, particle swarm optimization and an ensemble learning method collaborate for feature selection and cancer classification. As a result, the analysis indicates the effectiveness of the proposed method for cancer classification based on microarray datasets, and in terms of accuracy, the performance outcomes are 100%, 92.86%, 86.36%, 100%, 85.71% for leukemia, colon, breast cancer, ovarian, and central nervous system, respectively, which overcome most of the state-of-the-art methods and also dominance on the baseline ensemble method with 12% enhancement.