Time-series Forecast of the COVID-19 Pandemic Using Auto Recurrent Linear Regression

Abstract

Since the beginning of 2020, the COVID-19 pandemic has severely affected the economy and lifestyle of people. Various data analytics have been performed on data obtained from various sources. These analytics include symptom prediction, time-series forecasting, and impact analyses. Forecasting the end of the pandemic remains a challenge for many countries. Time-series forecasting models have been proposed for various applications. However, a nonseasonal and nonstationary forecasting method is required for predicting the progression of the pandemic. An auto regressive linear regression algorithm was proposed using the COVID-19 data of a certain geography. The results of the proposed methodology are convincing when compared with the nonseasonal and nonstationary existing methodologies, such as linear regression and exponential smoothing variants. The standard deviation and root mean square error of the proposed method were 430.22 and 0.31, respectively, for active cases, and 27.01 and 0.77 for the rate of transmission with positive skew and platykurtic trend, respectively.