Novel Coronavirus (COVID-19) Prediction using Deep Learning Model with Improved Meta-Heuristic Optimization Approach

This research develops a deep learning-based covid-19 prediction model by going through four essential phases. The obtained raw data is first submitted to the pre-processing phase to enhance the data quality and improvement in prediction accuracy of the proposed model. The data normalization and data cleaning are undergone in the pre-processing phase. The most important aspects, such as statistical features (Mean, Median, Min-Max and Standard Deviation), Exponential Moving Average (EMA), and Relative Strength Index (RSI), are retrieved from the pre-processed data. Self Improved Standard Spider Monkey Optimization algorithm (SI-SMO) is used to select the most optimal features. Following that, a new improved deep learning model, the optimized Deep Belief Network, is used to simulate the covid-19 prediction phase (DBN). DBN's weight is fine-tuned by the proposed SI-SMO algorithm. This parametric tuning of the model enriches its performance in predicting the disease. The outcome from optimized DBN portrays about the presence/ absence of covid-19. Accordingly, the accuracy value accomplished by the proposed model is 0.951001582 at LP=60, which is 16.6%,19%,15.4%,14%, 2.6% and 5.3% better than the traditional models like SVM, RF, CNN, NB, WOA+DBN, EHO+DBN, MFO+DBN and SMO+DBN, respectively.