TIME SERIES ANALYSIS FOR MODELING AND PREDICTING CONFIRMED CASES OF INFLUENZA A IN ALGERIA

Abstract

Influenza A is a subtype of the influenza virus that primarily infects birds and mammals, causing respiratory illness. It is characterized by its ability to mutate rapidly, leading to various strains and occasional pandemics. Objective. This paper is dedicated to studying the distribution behavior and predicting confirmed cases of Influenza A within the Algerian context, a highly infectious dis- ease that causes widespread illness and deaths both in Algeria and globally. Materials and methods. To predict confirmed cases of Influenza A, we implemented several statistical models, including ARIMA, Seasonal ARIMA (SARIMA), ETS, BATS, and the machine learning technique RNN, which is widely recognized in the literature. We then conducted a comparative study using performance measures to evaluate these models. Results. We used RMSE to determine the best-performing model. Our findings indicate that RNN outperformed the others due to its ability to handle complex patterns, including seasonal components and memory. SARIMA and BATS also performed well, thanks to their capacity to manage seasonal patterns. In contrast, ARIMA and ETS showed the poorest performance. Conclusion. This study employed a comprehensive approach to develop a model for predicting confirmed cases of Influenza A in Algeria. The results enhance our understanding of the potential future behavior of this disease and contribute to effective risk management strategies.