Final Analysis

Model Champion: Random Forest

Our team, ultimately, decided that the Random Forest class using the multi-outcome data set would be the best model to help hospitals predict potential future deaths. The primary reasons for this is that with an 87% degree of certainty we could predict bands of deaths due to COVID-19 and there is a high degree of explaining the model outcomes with the attributes the model selected.

This would help with hospitals ability to prepare in advance for critical resources like staffing, supplies, and beds depending on which bin the model predicted would happen 10 days in the future.


Champion Model Summary

Champion Model Approach- Used the Google Trend data with Binary outcomes is split into a standard 80/20 test/train split with a random state of 42.
Standard Scaling is also applied due to the larger number included in the Stocks Data Set. The model itself uses a random state of 1 with 500 estimators and yielded a Testing score of 0.876.

Champion Model Findings- One of the most interesting findings is that the Current[ly] Hospitalized COVID-19 Patients was again the most important feature . The second most important feature is “covid symptoms”.
Top 10 features were from the Google Trend data set: can dogs get covid, lake, quarantine, stimulus, lungs. Top 10 features from the Stock Data Set are the Open & Close Price for Moderna Inc (MRNA).
Rounding out the top ten features is the Total Doses Administered. The least important features were also from the Google Trend data (anti-depressant, anxiety, biden, gig economy).

Model Limitations and Acknowledgments

Limitations


While our initial model findings and predictions are promising, limitations exist with scaling the prediction to end users; with end users being a hospital or organization that would be able to use the prediction. This is due to, at least in part, to the original data used for defining outcomes from the CDC being aggregated at the national level.

In order to make our predictions more useful we would need to run our model(s) against localized or regional outcome data sets, and compare the predictions. This could potential be done by partnering with NGO's or Health Care providers with more granular data. This could also lead to needing more granular Google Trend Search data as well as other input data sets which could prove challenging.