Abstract
With the rapid development of health information systems in recent years covering big data trends, clinical information is being adapted to understand important health trends and provide timely prophylactic treatment. However, the current plan fails to maintain a highly organized analysis and processing of health data. In addition, it collects issues that play an important role in determining quality. In this article, we propose a hybrid optimization based learning technique for multi-disease analytics (HOL-MD) from healthcare big data using optimal pre-processing, clustering, and classifier. First, we introduce a capuchin search based optimization algorithm for preprocessing which removes the unwanted artifacts to enhance the detection accuracy. Second, we introduce a modified Harris Hawks optimization based clustering (MHHOC) technique used to select optimal features among multiple features which discovers the subgroups and reduce the dimensionality issues. We evaluated the performance of proposed HOL-MD technique using standard US healthcare organization dataset, SUSY and HIGGS datasets. Finally, we proved the effectiveness of proposed RTL-DNN techniques in terms of 11.5%, 9.3%, 2.75%, and 3.45% higher than the existing state-of-art techniques in terms of accuracy, precision, recall, and F-measure respectively.