Feature Selection With Novel Mutual Information and Binary Grey Wolf Waterfall Model

Abstract

This article aims to identify a way to predict cancer by analyzing gene expression data from microarrays. The focus is on selecting specific features through metaheuristic search algorithms that can help determine the optimal global and local features using population and neighborhood-based methods. Essentially, the goal is to use advanced technology to identify potential biomarkers for cancer prediction, through a novel computer-aided diagnostic tool for the classification of cancer samples using gene expression data. The model is known as JMR-CR with waterfall GWO and operates in two distinct phases. In the initial phase, the JMI-CR algorithm is employed to select the most relevant features from the dataset by utilizing a novel mutual information technique called joint mutual information. In the second phase, the waterfall grey wolf optimization algorithm is used to identify the optimal features. To assess the performance of the proposed model is evaluated using two classification algorithms, namely support vector machine (SVM) and K nearest neighbor (KNN). The proposed model offers several advantages. It addresses the challenges posed by higher dimensionality and class imbalance problems by utilizing the waterfall GWO model, resulting in increased classification accuracy. The model is tested on various cancer microarray gene expression datasets, and the experimental results demonstrate that the proposed hybrid model outperforms other existing models in terms of generalization performance and testing accuracy.