This message was deleted.

You could, but realistically it becomes harder to deploy also and there is a high chance of overfitting when you ensemble like this. I guess the question is what is the use case and does a difference of 0.1 or 0.2 accuracy have significant impact?

Hey @marah othman I wouldn't take the average but the median. I don't know what kind of models you're planning to use, but here's a good example of a combination of univariate models that placed 6th in the M4 Competition. All the models described in that paper are implemented in StatsForecast.

@Kevin Kho i am not sure if you understand my question i am asking if i can get the predcitions after train all models one by one and sum this prediction by average or meadian to benfit from all results

I think I understand. I’m wondering how much the performance of your predictions increases by if you do this, and if it’s worth it.