学术文献库

The coronavirus disease 2019 (COVID-19) has become a public health emergency of international concern affecting more than 200 countries and territories worldwide. As of September 30, 2020, it has caused a pandemic outbreak with more than 33 million confirmed infections and more than 1 million reported deaths worldwide. Several statistical, machine learning, and hybrid models have previously tried to forecast COVID-19 confirmed cases for profoundly affected countries. Due to extreme uncertainty and nonstationarity in the time series data, forecasting of COVID-19 confirmed cases has become a very challenging job. For univariate time series forecasting, there are various statistical and machine learning models available in the literature. But, epidemic forecasting has a dubious track record. Its failures became more prominent due to insufficient data input, flaws in modeling assumptions, high sensitivity of estimates, lack of incorporation of epidemiological features, inadequate past evidence on effects of available interventions, lack of transparency, errors, lack of determinacy, and lack of expertise in crucial disciplines. This chapter focuses on assessing different short-term forecasting models that can forecast the daily COVID-19 cases for various countries. In the form of an empirical study on forecasting accuracy, this chapter provides evidence to show that there is no universal method available that can accurately forecast pandemic data. Still, forecasters' predictions are useful for the effective allocation of healthcare resources and will act as an early-warning system for government policymakers.