Applying Deep Reinforcement Learning Techniques for Covid-19 Case Prediction

Although restrictions and weariness against COVID-19 have eased, the pandemic is still very much existent. Additionally, we may encounter similar viruses in the near future. Therefore, we need a reliable and reproducible system that may detect and predict these changes to create better countermeasures. Machine Learning offers a solution by efficiently solving complex models and trends, enhancing prediction accuracy. Furthermore, Reinforcement Learning (RL) has demonstrated an outstanding ability to learn and outperform other algorithms and man-made applications. Combined with Deep Learning (DL), Deep Reinforcement Learning (DRL) achieves better computation power and accuracy. Although very few have explored the application of DRL to predict COVID-19 cases, the many computational and performance-enhancing benefits of such methods pose great potential for solving similarly complex environments. In this paper, we design a naive DRL algorithm to analyze, forecast, and predict COVID-19 cases under varying controls to construct efficient mitigation protocols for future applications.