Modeling the Spread of COVID-19 by Leveraging Machine and Deep Learning Models

Corona Virus disease 2019 (COVID-19) has caused a worldwide pandemic of cough, fever, headache, body aches, and respiratory ailments. COVID-19 has now become a severe disease and one of the leading causes of death globally. Modeling and prediction of COVID-19 have become inevitable as it has affected people worldwide. With the availability of a large-scale universal COVID-19 dataset, machine learning (ML) techniques and algorithms occur to be the best choice for the analysis, modeling, and forecasting of this disease. In this research study, we used one deep learning algorithm called Artificial Neural Network (ANN) and several ML algorithms such as Support Vector Machine (SVM), polynomial regression, and Bayesian ridge regression (BRR) modeling for analysis, modeling, and spread prediction of COVID-19. COVID-19 dataset, maintained and updated by JOHNS HOPKINS UNIVERSITY was used for ML models training, testing, and modeling. The cost and error generated during ANN training process was reduced using technique called back propagation which dynamically adjust the synapses weights to perform better predictions. The ANN architecture included one input layer with 441 neurons, 4 hidden layers each have 90 neurons and one output layer. ANN along with other ML algorithms were trained to model the prediction of COVID-19 spread for the next 10 days. Experimental results showed that BRR technique overall performed better prediction of COVID-19 for the next 10 days. The modeling of infectious diseases can help relevant countries to take the necessary steps and make timely decisions.