A novel deep learning-based multi-model ensemble method for the prediction of neuromuscular disorders

Neuromuscular disorder is a complex progressive health problem which results in muscle weakness and fatigue. In recent years, with emergence and development of machine learning- and sequencing-driven technologies, the prediction of neuromuscular disorders could be made on the basis of gene expression for accurate diagnosis of disease. The intent is to correctly distinguish the patients affected from neuromuscular disorder from the healthy one with the help of various classification methods used in machine learning. In this paper, we proposed a novel feature selection method which applies deep learning method for grouping the outputs generated through various classifiers. The feature selection is performed on the basis of integrated Bhattacharya coefficient and genetic algorithm (GA) where fitness is computed on the basis of ensemble outputs of various classifiers which is performed using deep learning methods. The Bhattacharya coefficient computed the most effective gene subset; then, the most discriminative gene subset will be formulated using GA. The proposed integrated deep learning multi-model ensemble method was applied on two commercially available neuromuscular disorder datasets. The obtained results encouraged that the proposed integrated approach enhances the prediction accuracy of neuromuscular disorders as compared with different datasets and other classifier algorithms. The proposed deep learning-driven ensemble method provides more accurate and effective results for neuromuscular disorder prediction and classification with the help of distinguished classifiers.