Spatio-temporal covid-19 prediction using hybrid gated attention network-RNN model

Abstract

As we all know covid19 pandemic heated us deeply. The entire world goes down in lockdown many times. Many major countries’ economies collapsed, and the healthcare system collapsed due to the major number of covid 19 patients. However, many countries or states are proposed an international travel ban in early 2020, so there are many reasons to transmissions would to occur due to domestic travel in-between states or cities. Different states or cities have different populations, some cities have dense populations and some have fewer populations. While extensive research has analyzed the issue on a large scale to predict the pandemic, here we focus on entire countries and states, mainly comprising states and cities, for a more accurate prediction of the number of cases (like new cases, death cases, etc.) we use different country or state claim data, demographic and geographic proximity between location and divide them based on populations, and integrate pandemic dynamic into the deep learning model. We perform an attention network model to predict the pandemic data. It uses a graph attention network to dynamically predict cases for a certain period in a country or state. By using real-world data and case counts, this model accurately predicts covid19 status and helps us to prevent pandemics in the future.