Surveil and Prediction of Pandemic Disease by Fine-Tuning Hyperparameters in Deep Learning

Abstract

The major issue that is being faced by the medical professional in the world of health care is the accurate diagnosis of conditions and diseases of the affected patient. In the current scenario, Coronavirus (COVID-19) infections are major challenging in the health care sector. The medical diagnosis for COVID-19 prediction is done with the support of X-ray images. Detection of COVID-19 with X-ray images is an effective method but it is critical for diagnosis, evaluation, and medication. Since the X-ray images are of low contrast, the accuracy will be low while using the traditional feature extraction in machine learning. The proposed method uses a novel technique by combining multiple CNN (Convolutional Neural network) for classification. The samples for multiple CNN (Convolutional Neural Network) are divided into n-subsets by using the bootstrap sampling method. The n-subset samples are finely tuned with Adaptive momentum technique (Adam) optimizer, the output of this layer is fed into CNN (Convolutional Neural Network) model as input, still hyper-tuning at each layer with Bayesian optimizer. This technique avoids de-noising, and features are also enhanced when compared to conventional methods. Experimental results of this technique provide an accuracy of 97.35% for the prediction of COVID-19 from X-ray images.