Model 5
Random Forest model with uncertainty computed with conformal prediction
| Author | Eduardo Correa Araujo |
|---|---|
| Disease | Dengue |
| Administrative Level | Municipality |
| Time Resolution | Week |
| Implementation Language | Python |
Description:
In this repo, in the path /models/gbt it's implemented a random forest regressor model that
based on the last four weeks of data (cases and climate variables), compute the cases in the fourth week ahead.
The predictions for multiple times are obtained in a rolling window fashion, i.e.,
the historical data window is moved forward one week at a time, predicting the next fourth week at each step.
The confidence interval of the predictions are computed using the conformal prediction.