Statistical machine and deep learning methods for forecasting of Covid-19

Abstract

Abstract Since the outbreak of the novel coronavirus, Covid-19 has continuously spread across the globe briskly. Countries have undertaken different types of measures to blunt this spread varying from lockdowns to curfews to social distancing to compulsory wearing of protective kits, which has been sporadically fruitful. However, despite these stringent measures, which have their own pitfalls, scientists across the globe have been struggling to develop a suitable mathematical model that could depict the existing disease spreading pattern and also predict a trend of numbers in the forthcoming months or years. In this paper, popularly used mathematical models including Polynomial Regression, Auto Regressive Integrated Moving Average (ARIMA) and Deep learning techniques such as Recurrent Neural Network (RNN) have been explored for 5 countries badly affected by this virus. The models were tested from 16th May, 2020 till 22nd May, 2020 and used for predicting future cases and deaths from 23rd May, 2020 to 30th June, 2020. The current research primarily focuses on forecasting the behaviour of total confirmed cases and deaths in each country and further analysing the performance parameters such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE) and Mean Absolute Percentage Error (MAPE). It has been observed that the polynomial regression model provides a best fit solution at par with actual numbers of confirmed and death cases for India by producing minimum RMSE and MAPE. For South Korea and Italy, the ARIMA and RNN models have shown fidelity with actual numbers. RNN model has shown conformity with US numbers while ARIMA model has found closeness to United Kingdom data. The purpose to perform data analysis is to measure the performance metrics by using different techniques and depict the pattern for each country. Furthermore, the paper also highlights the future predictions for every country to control the spread of disease, save lives, avoid health systems breakdowns and benefit the researchers in this field.