In collaboration with Gavin Hudgeons and Southern Methodist University

Utilizing Nationwide and Arizona-specific data captured by Johns Hopkins and the Covid Tracking Project, we calculated the rate of positive test percentage and used data dating from May – July to predict short term and longterm performance forecasts moving forward. Below is a link to the videos to see our presentation of the results.

For the US Data, after accounting for the change in trend to the data, we can forecast using a stationary model with an average square error rate of 3.54e-05.

For Arizona-Specific data, we are able to forecast to the 0.001 level of Average Square Error, which were both strong overall when differencing to the trend.

For the US forecast, this would suggest positive percentage rate is stabilizing towards a level of zero as it continues to approach it through the end of July’s data going into August. For Arizona, who is still in its apex of positive percentage at the time of this project, we’re seeing some cyclical spikes in positive percentage which will continue for the next 20 days based on our forecast model.

Below is a presentation

EDA:

Final Presentation

Other links

Github Link