Authors: Anshuman S, Arvind B, Bhavay A. Affilation: Georgia Institute of Technology, Atlanta.
Webiste link: https://sites.google.com/view/covid-19-dssta
Project presentation: https://gtvault-my.sharepoint.com/:p:/g/personal/abangaru3_gatech_edu/Ec3C_uayF9xDomH8UsMb8TMBF6yJSes1pcTUfUVNj3eqAQ?e=3F1tfd
https://github.com/anshumansinha16/EPI_final_Submission/blob/main/EPI_final_Abstract_mid.pdf
Deep learning, Social-Graphs, Google search trends, Covid-19, , Epidemiology, Time- series, Disease forecasting, Pytorch-forceast, SIR-Network
The use of hospital resources and the development of management plans to best manage infected patients depend on accurate forecast- ing of COVID-19 cases. However, monitoring sensors is an excellent method to measure the spread of the disease; it is an expensive task and has various privacy and ethical issues. A low-cost alternative to sensors in monitoring the sensors can be leveraging search trends on disease symptoms. It is not far-fetched to think that people with such symptoms would google them to get more information and potential remedies or cures. If this assumption is experimentally validated, it becomes possible to forecast the disease by forecasting symptoms related to Covid-19. In this work, we look at how the dis- ease forecasting problem has been previously approached and how search trends data can be incorporated into it. We have explored the possibility of using search trends as a proxy to actual disease cases with the help of SIR-Network model. We extend the prediction of cases to inter-regional space as well, with spatio-temporal predic- tions with the help of search trends. We also looked at predicting pharmacological demands such as vaccines, and miss-information of disease through search trend results.