Enhancing Cancer Classification Through the Development of a Fuzzy Gene Selection-Wrapper Plus Method

Deep Learning (DL) approaches have made sub-stantial advances in developing classification models in most disciplines, in terms of both accuracy and speed. However, analyzing gene expression data for cancer classification remains challenging due to the nature of the complexity of available datasets, which are characterized by high dimensionality (a small number of samples and a large number of features). We developed a new fuzzy gene selection wrapper plus (FGSWP) to select the most significant genes with the goal of reducing the complexity of cancer classification and the dimensionality of the datasets. FGSWP has shown its efficacy by lowering the number of genes and enhancing classifier model performance. Wrapper techniques are not suitable for use with high-dimensional data, hence fuzzy gene selection is used as the initial phase for reducing the number of genes before utilizing wrapper techniques. The accuracy attained in all datasets used ranged from ${85.3{\%}}$ to 99.3% when FGSWP and Multilayer Perceptron (MLP) were employed together.