Table 6.

A relative comparison of AI techniques used in tracking and prediction

Authors	Year	Objective	AI technique	Merit	Demerits	Results
Tuli et al. [154]	2020	The authors proposed a model which predicted daily change in cases as well as classifying them using AI Technique	Inverse Weibull function	Shows better prediction result compared to baseline Gaussian model	Major factors like age difference, population density and temperature were not considered	$R^2$ value is 0.98
Gupta et al. [46]	2020	The authors proposed an AI-based ARIMA model for forecasting the number of cases in India and predicting the future trends	ARIMA Model	Can aid scientists as well as the government agencies to make their plans	The model is not capable to show the output of different regions	Accuracy is around 60–70%
Malik et al. [86]	2020	The authors presented an approach to predict the number of cases from early infection dynamics	Gradient Descent	Easy to implement and works successfully with aggregate data	Changing public health protocols can skew the results	$R^2$ value is 0.57
Sina et al. [9]	2020	The authors proposed a ML-based SIER model to predict the spread of virus via human intervention	ANN	Shows more accuracy where the government delayed the containment	Skewed results in regions where people are willing to quarantine themselves and restrict social contact	Root mean Square came out to be 1028.98 for logistic model in Italy
Bengio et al. [15]	2020	The authors proposed a model to curb the exponential growth of the COVID-19 via contact tracing	ML	Can give information about the individual risk in a detailed manner	Security breaches can violate the privacy of the users
Ahmada et al. [1]	2020	The authors identified the number of confirmed cases using AI Technique	Multivariate linear regression	Successfully shows real-time outbreak even with small datasets	Lack of historic data as previous pandemics were different and their results cannot be used here	Models should be based on multiple types of data
Liu et al. [83]	2020	The authors proposed a model for real-time prediction of COVID-19 based on Internet searches and news alerts using AI Techniques	LASSO multi-variable regularized linear model	Able to predict output 2 days prior	Depends upon the geographical location	Normalized root mean square value was greater than 1
Ayyoubz-adeh et al. [11]	2020	The authors analyzed the Google trends and predict the incidence of COVID-19 at different region	LSTM	Very low training errors are discovered	Limited access of Google search at some regions	Root mean square value for LSTM model came out to be 27.187