Table 3.

Different time series regression models for different infectious diseases.

S. no.	References	Conclusion drawn
1	Zhang et al. (15)	In this study, the authors have presented a complete analysis of different predicting methods based on the monthly infection spread data of typhoid fever. The seasonal autoregressive integrated moving average (SARIMA) model and three different models inspired by neural networks, namely, backpropagation neural networks (BPNN), radial basis function neural networks (RBFNN), and Elman recurrent neural networks (ERNN) were compared. The dissimilarities, pros, and cons, between the two models. The evaluation was based on three metrics: mean absolute error (MAE), mean absolute percentage error (MAPE) and mean square error (MSE). The results showed that RBFNN obtained the smallest MAE, MAPE, and MSE in both the modeling and forecasting processes. Ultimately, it was suggested to use the RBFNN method for better explanation and prediction of typhoid fever infection spread.
2	Zhang et al. (74)	In this work, nine types of infections were compared based on the efficiency of four-time series methods, regression and exponential smoothing, ARIMA, and support vector machine (SVM). The performances were evaluated based on three metrics: mean absolute error (MAE), mean absolute percentage error (MAPE) and mean square error (MSE). The robustness of the statistical models in predicting the potential spread of the infections showed their good application in epidemiological surveillance and found that no single method is completely superior to the others but support vector machine-based methods are proven better than the ARIMA models and decomposition methods in most of the cases.
3	Imai et al. (75)	In this study, time series regression was applied to evaluate the short-term associations of air pollution and weather with mortality or morbidity of infectious diseases. They used different approaches, including mathematical modeling, wavelet analysis, and ARIMA models. They concluded that the time series regression can be used to investigate the dependence of infectious diseases on weather, but may need modifying to allow for features specific to this context.
4	Song et al. (76)	The authors compiled monthly data of influenza infections from all provinces and autonomous regions in mainland China and applied the time series analysis to construct an ARIMA model. They have evaluated the goodness of fit through Autocorrelation function (ACF), partial autocorrelation function (PACF), and automatic model selection was to determine the order of the model parameters. It is conceivable that SARIMA is the best time series model for the prediction of influenza infection spread.
5	Sarkar and Chatterjee (77)	The authors have applied different time series models to analyze and forecast financial data as well as epidemiological data of malaria infection dissemination. They have studied epidemiological data of malaria using three-time series models, namely Auto-Regressive Integrated Moving Average (ARIMA), Generalized Auto-Regressive Conditional Heteroskedastic (GARCH), and Random Walk. They have shown a good fit of models on the data and provided the best forecast for future infection spread. As far as future prevalence pattern is concerned, the prediction of these models may help researchers and public health professionals to design control programs for malaria.
6	Chae et al. (78)	The authors studied the prediction of infections by optimizing the parameters of deep learning algorithms while considering big data including social media data. The performance of the deep neural network (DNN) and long-short term memory (LSTM) learning models were compared with the autoregressive integrated moving average (ARIMA) when predicting three infections for 1 week time into the future. They have shown that the DNN and LSTM models perform better than ARIMA. The DNN model performed stably and the LSTM model was more accurate when infections were spreading.
7	Tapak et al. (79)	The author analyzed the correctness of support vector machine, artificial neural network, and random-forest time series models in influenza-like illness (ILI) modeling and infection detection. Different models were applied to a data set of weekly ILI cases data in Iran. To judge the robustness of the models, the root means square errors (RMSE), mean absolute errors (MAE), and intra-class correlation coefficient (ICC) calculations were used as testing criteria. It was indicated that the random-forest time series model worked better in comparison to the rest three methods. The outcome depicted that the used time series models had excellent performance suggesting these could be effectively applied for predicting weekly ILI infections and endemics.
8	Chaurasia and Pal (80)	In this work, the authors have analyzed the number of cases, deaths, and recovery cases in the case of COVID-19 worldwide within a specific period. They have used several prediction techniques: naive method, simple average, moving average, single exponential smoothing, Holt linear trend method, Holt-Winters method, and ARIMA, for comparison, and how these methods improve the Root mean square error score. They concluded that the naive method is best in comparison to other used methods.
9	Rahmadani and Lee (81)	The authors suggested a hybrid deep learning framework using the meta-population model and long and short term model (LSTM) for the prediction of the COVID-19 dissemination. They expanded the susceptible–exposed–infected–recovered compartment model by taking into account the human mobility among a number of regions. They used the meta-population model to incorporate with deep learning models to estimate the parameters of the combined hybrid model. They have compared the suggested hybrid deep learning framework with other estimation methods for the prediction of COVID-19 spread patterns and have shown improvement over previously presented methods.
10	Kalantari (82)	The author used the singular spectrum analysis (SSA) method for the prediction of the number of daily confirmed infection cases, deaths, and recoveries caused by COVID-19. It was analyzed using SSA method with the other commonly used time series predicting techniques including ARIMA, fractional ARIMA, exponential smoothing, TBATS, and neural network autoregression (NNAR) on the basis of fitting measure root mean squared error (RMSE). It was shown that the SSA technique is best for predicting the number of daily confirmed infection cases, deaths, and recoveries caused by COVID-19 among the studied models.
11	Satrio et al. (83)	The authors utilized the machine learning model for predicting the spread of COVID-19 in Indonesia. They have also attempted to estimate a time line for the return of the normalcy. They have utilized PROPHET forecasting model as well as ARIMA to see their robustness and accuracy for the confirmed new infection cases, deaths, and recovered numbers. They have shown that PROPHET performs better than ARIMA model on the analyzed data set.
12	Beneditto et al. (83)	The authors utilized the Machine Learning model to forecast the trend of the disease in Indonesia with finding out the approximation when normality will return. This study used Facebook's Prophet Forecasting Model and ARIMA Forecasting Model to compare their performance and accuracy on a dataset containing the confirmed cases, deaths, and recovered numbers, obtained from the Kaggle website. The prediction models are then compared to the last 2 weeks of the actual data to measure their performance against each other. The result showed that Prophet has predicted the outcomes better than ARIMA, despite it being further from the actual data the more days it predicts.