Hi @emercado777,
I went and split some of the larger slides up and included them as a reply in the project update forum post for my submission (Power BI Challenge 11 - Covid 19 Reporting from Brad). The demographic data was really what drew me into focusing on a specific state so that I could really do it justice.
And I didn’t really consider adding a forecast of new cases as a variable, but as one of the charts I did. That’s probably the most common use of the SEIR model and almost every forecast put out by a health organization uses a variation of the SEIR model. I take it back, I did consider using a LSTM model but the main drawback to that is that LSTM models for viral infections generally just focus on the number of infected and ignored other factors that truly have an impact on cases. It would be interesting to create a hybrid approach that uses the two together. It has actually been very interesting to see how these models have changed due to COVID.
Most of the models used were originally created fairly early into the pandemic and were created from much smaller/shorter outbreaks/pandemics. This is why many of these models only forecast 7-14 days out, they never had the such vast amounts of data for a single pandemic nor has it been this wide reaching so the models are subject to overfitting. A number of variants of the SEIR model have been developed as a result and some of them are truly incredible. There’s one in particular out of India that I really wanted to try but is much more complex and I didn’t have time to implement.
If you can’t tell, I get pretty excited about machine learning and these types of models haha