An hybrid technique for optimized clustering of EHR using binary particle swarm and constrained optimization for better performance in prediction of cardiovascular diseases

The significant adoption of Electronic Health Records (EHR) in healthcare has furnished large new quantities of information for statistical machine gaining knowledge of researchers in their efforts to version and expects affected person health popularity, doubtlessly permitting novel advances in treatment. Unsupervised system learning is the project of studying styles in facts where no labels are present. In comparison to loads of optimization problems, an most beneficial clustering end result does not exist. One-of-a-kind algorithms with special parameters produce special clusters, and none can be proved to be the quality answer given that numerous good walls of the records might be found. In the previous work, a novel Two-fold clustering technique which uses the Long Short Term Memory (LSTM) technique (TFC: LSTM) for the prediction of Cardiovascular Disease (CVD) was proposed. The proposed model was fond to be experimentally efficient; however when applied to large EHR data, the model suffered from optimization issues on the number of clusters formed and time complexity. In order to overcome the drawbacks, this paper proposes a hybrid method of optimization using the Binary Particle Swarm (BPS) and Constrained Optimization (CO) for optimizing the number of clusters produced and to increase the efficiency in terms of decreasing the time complexity.