Atom Taylor Bird Swarm algorithm-based deep belief network for incremental classification using medical data

Abstract

Medical data classification is considered as a challenging and complex task in the field of medical informatics. Various medical data classification methods are developed in the existing research works, but to achieve higher performance in terms of classification accuracy result in a great challenge in the medical sector due to the presence of missing values, uncertainty and redundant attributes. Hence, an effective and optimal data classification method named Atom Taylor Bird Swarm Algorithm-based Deep Belief Network (Atom Taylor BSA-based DBN) is proposed in this research to perform incremental classification using medical data. The proposed Atom Taylor BSA is designed by integrating the atom search optimization (ASO) with the Taylor Bird Swarm algorithm (Taylor BSA), which is the development of the Taylor series with BSA. The DBN classifier effectively performs the incremental classification using medical data with its associated neurons and generates optimal result based on the fitness measure. Accordingly, the selected features enable the classifier to increase the performance of classification accuracy. However, the proposed Atom Taylor BSA-based DBN obtained better performance using the metrics, like specificity, sensitivity, and accuracy with the values of 0.8307, 0.9074, and 0.8804, using Hungarian dataset.