Improved gene expression diagnosis via cascade entropy-fisher score and ensemble classifiers.

Feature selection is an important technique used in bioinformatics modeling to reduce the dimensionality of high-dimensional data. However, filter-based approaches that have shown better performance often depend on specific measurement methods, which can limit their effectiveness. To address this problem, this paper proposes a novel cascade feature selection approach, named the cascade entropy-fisher score (CEFS), that combines entropy score (ES)-based and Fisher score (FS)-based feature selection. CEFS involves a two-step process where in the first step, the entropy of each gene in the dataset is calculated to measure the uncertainty associated with its expression levels across different samples. In the second step, the Fisher score is computed to measure the extent to which the gene's expression levels differ between classes of samples. CEFS has been shown to outperform other methods in identifying disease-specific genes in gene expression datasets, making it a promising tool for disease diagnosis and prognosis. The proposed method was evaluated on biomedical datasets, and its effectiveness was measured in terms of accuracy, sensitivity, specificity, and area under the curve (AUC). The results showed that CEFS has comparable performance to state-of-the-art feature selection methods in the literature. Additionally, the selected features were fed to an ensemble of three classifiers, including support vector machine (SVM), k-nearest neighbor (k-NN), and decision tree (DT), to evaluate performance in the classification stage. The ensemble approach is based on majority voting, which aggregates the outputs of the individual classifiers to determine the final label. The results demonstrate the potential of CEFS in machine learning applications, particularly in the context of disease diagnosis and prognosis. [ABSTRACT FROM AUTHOR]