چكيده به لاتين
Given the circumstances arising from the COVID-19 pandemic in Iran, it is essential to predict the spread and impact of COVID-19. Considering that COVID-19 is a complex and widespread disease with significant effects on societies and healthcare systems, accurate predictions can help in proper planning. By estimating the number of new cases or deaths, healthcare resources and equipment can be effectively managed, and the supply chain and distribution can be improved. Additionally, forecasting the trend of the virus allows for the implementation of effective measures to reduce its spread and manage responses to future waves of the disease. Overall, accurate prediction of COVID-19 enables decision-makers to make crucial decisions based on reliable information and optimally allocate resources and measures. In this study, our main objective was to develop an improved approach for predicting COVID-19 cases in Iran. We started by conducting a comprehensive analysis of various factors that contribute to the spread of COVID-19. These factors included the transmission rates of different viral variants, the efficacy of vaccines against different variants, and the impact of climate on the transmission rate. After carefully examining these characteristics, our next step was to assess the practicality and accuracy of different prediction techniques. We evaluated several commonly used methods, including ARIMA, discrete logistic model, SIRDH model, and Bi-LSTM networks. By analyzing their ease of use and prediction accuracy, we aimed to identify the most suitable technique for our purposes. Based on our analysis, we found that the implementation of the logistic model generally demonstrated higher feasibility and simplicity. By accurately estimating the model's parameters, we were able to achieve acceptable prediction results. As a result, we devised two methodologies to forecast the number of COVID-19 mortalities during each surge in Iran. In our initial approach, we began by estimating the parameters of the logistic model. To enhance the accuracy of the primary parameter, we incorporated predictions from other models and employed a two-layer neural network. However, we encountered some complexities with this method, prompting us to introduce another approach. As a result, we have chosen to streamline our process by relying solely on the available data and the logistic model's prediction based on initial guesses.to summarize, our study involved an extensive analysis of various prediction techniques to identify the most user-friendly and accurate model. We then applied the logistic model, refining its parameters with the help of other models and a neural network. Ultimately, we simplified our approach by utilizing the available data for predicting COVID-19 mortalities in each surge.
