Deep enriched salp swarm optimization based bidirectional -long short term memory model for healthcare monitoring system in big data

Abstract

The rapid development of communication technologies and expert systems have resulted in a large volume of medical data. Big data such as clinical data, omics data, and electronic health data are difficult to manage in real-time due to noise, large size, different formats, missing values and large features. Hence, it is more difficult for the health monitoring system to extract the correct information. Low quality and noisy data can lead to unnecessary treatment. To overcome these issues, we proposed Enriched Salp Swarm Optimization based Bidirectional Long Short Term Memory (ESSOBiLSTM) to monitor health. This method consists of four layers, such as the data collection layer, data storage layer, data analytics, and presentation layer. The initial layer handles a variety of information from main sources: wearable sensor devices (WSD), social network data, and medical records (MR). The second layer stores all the collected data from WSD, MR, and social network data to the cloud server through the wireless network. The proposed framework for performing big data analytics steps like preprocessing, filtering, dimensionality reduction, and classification is performed in the third layer. In the final layer, the doctor analyzes the patient's condition based on the classification results of the enriched SSO-BiLSTM. Based on the evaluation report, the proposed ESSOBiLSTM gives an accuracy of 85%, precision of 80%, RMSE of 0.6, MAE of 0.58, recall of 85% and F-measure of 79%. As a result, ESSOBiLSTM has proven to be more effective in monitoring health in large datasets.