COX Regressive Winsorized Correlated Convolutional Deep Belief Boltzmann Network for Covid-19 Prediction with Big Data Dr. K.Sankar 1 , Dr. Divya Rohatgi 2 and S.Balakrishna Reddy 3 1, 3 CVR College of Engineering, Dept. of CSE, Hyderabad. India Email: sankarkrish@cvr.ac.in, sama.balakrishnareddy@gmail.com 2 Amity University, Dept. of CSE, Mumbai, India Email: divi.rohatgi@gmail.com Abstract—Big data analytics in the health care industry is a very promising method of integrating, discovering, and analyzing huge volume of complex heterogeneous data. Some of the techniques were not well suited for medical data classification with minimum time consumption. To improve disease prediction accuracy, a novel technique Cox Regressive Winsorized Correlated Convolutional Deep Belief Boltzmann Network (CRWCCDBBN) is introduced. The proposed CRWCCDBBN technique uses multiple layers for performing the different processes like feature selection and classification. Initially, feature selection is performed using Cox Regression to minimize the dimensionality of the data by finding the Cox partial log-likelihood between the two features. After the feature selection, the classification is performed using Winsorized Correlation coefficient by analyzing the training and testing disease data. Based on the classification results, Covid 19 disease is correctly diagnosed with a minimum error by updating the weight value. Finally, the gradient descent first-order iterative function is used to find the local minimum error. Experimental evaluation of CRWCCDBBN technique is carried out using a Covid 19 Dataset with different performance metrics such as accuracy, precision, recall, f-measure, and prediction time with respect to a number of patient data. The observed result proves that the proposed CRWCCDBBN technique achieves higher prediction accuracy with a minimum time consumption than the state-of-the-art methods. Index Terms— Big Data, Convolutional Deep Belief Boltzmann Network, Cox Regression, feature selection, Winsorized Correlation coefficient, gradient descent first-order iterative function. I. INTRODUCTION A novel virus disease called COVID-19 caused in the world with various problems. But there are still limitations in the real-time detection of COVID-19 such as a large volume of datasets, imbalance classes, and a misclassification rate of models. Therefore, a novel deep learning-based model aims to investigate for effective detection of COVID-19. A hybrid deep learning prediction model CNN-LSTM was introduced in [1] to correctly forecast the COVID-19. However, it failed to test the algorithm on the large time-series datasets to determine suitability and correctness. Deep-LSTM ensemble model was introduced in [2] to predict the Covid-19. But it was not efficient to provide better results with minimum time. Supervised machine learning models were Grenze ID: 01.GIJET.9.1.547 © Grenze Scientific Society, 2023 Grenze International Journal of Engineering and Technology, Jan Issue