Closed andre-ls closed 4 years ago
Idea
Create a function that given the WF, Hour and Day of the Run, plot the weather features predictions for each NWP.
The idea here is to compare the predictions of each NWP and check if they are very different or not. If they're similar, it feels resonable to "ensemble" the NWP predictions by a simple method like mean.
The data presented by the challenge has a lot of weather predictions for a same feature on a same datetime. The idea here is to try to simplify the dataset before doing EDA, by makiing a kind of "Ensemble" of the NWP's predictions.
But the question is: Keep the features separated is important for the model or not?