Some models in the ensemble do not generalize well. For example, exponential models will likely not perform well at extrapolation, as they can lead to explosive predictions of production. To get around this, we could limit the "extrapolation range" of each of the models (i.e. exponential models can impute all interpolations but only extrapolations within one time step of an observed value, say).
Some models in the ensemble do not generalize well. For example, exponential models will likely not perform well at extrapolation, as they can lead to explosive predictions of production. To get around this, we could limit the "extrapolation range" of each of the models (i.e. exponential models can impute all interpolations but only extrapolations within one time step of an observed value, say).