In which areas/ problems stacked ensemble is useful compared to other models in a specific industry/ application. Commonly, either simple model such as linear regression is utilized for explanaibility while black-box models such as neural network were used in different cases. Where does stacked ensemble models stand?
To my knowledge there's no specific use-case for stacked ensemble models.
- There are technical constraints: by definition these models require training multiple individual models so the training must not be too intensive (e.g. one rarely use it with DL models which need very long training).
- There are some advantages in terms of performance and especially of stability/reliability of the results, since these models are statistically more robust to noise (i.e. less likely to ovefit).
From the point of view of explainability, ensemble models tend to be complex to explain since the explanation of a prediction involves multiple individual models. So it can be doable but it's not convenient at all.