SIRD-ABiGRU-AE: A Modified Compartmental Model with Attention-Driven BiGRU Autoencoder for COVID-19 Outbreak Prediction and Hospitalizations

Several epidemiological studies have been undertaken using a compartmental model to predict disease spread effectively. However, knowledge about the epidemiological cycle lacks existing techniques and fails to promote the vaccines and medications that the government issues to overcome the pandemic disease. Many researchers implemented a Susceptible-Infected-Recovered-Deceased (SIRD) based compartmental approach to determine the methods emphasized by the government to eradicate the spread of COVID-19. The traditional SIRD-based compartmental model produces high prediction errors and is time-consuming. Hence, this article presents a novel Deep Learning (DL) based Attention-driven bi-directional gated recurrent unit Autoencoder (A-Bi-GRU-AE) model, which is hybridized with the SIRD model to enhance the system performance. The proposed approach is implemented in the PYTHON platform, and the publicly available covid19Italy dataset is utilized for the experimental process. The proposed method obtains the overall predicted R2 of 0.97 and time complexity of 2634.01ms.