Butterfly Optimization and Neural Network Based Feature Selection and Classification of Gene Expression Data

Imagine a world where cancer could be detected at its earliest stages, paving the way for timely and effective treatment. Efforts to achieve this goal through innovative approaches have been ongoing. One promising solution is found at the intersection of deep learning and evolutionary algorithms. This paper introduces DenseNet, a novel approach that synergistically integrates the butterfly optimization algorithm (BOA) for feature selection and deep learning models like CNN for accurate and early classification of cancer types from gene expression data. BOA, inspired by butterfly foraging behavior, effectively identifies the most relevant gene features, enhancing model performance. The selected features are converted into a 2D image format and fed into the DenseNet architecture, a CNN designed to mitigate issues like the vanishing gradient problem.Additionally, the paper explores the application of other deep learning algorithms like convolutional neural networks (CNNs) and deep neural networks (DNNs) for gene expression data classification. The collective findings contribute valuable insights into accurate cancer classification and emphasize the potential of deep learning methodologies advancing bioinformatics research.