|
For a machine learning task I am trying to combine two models, but I am a bit puzzled on how I should generate the predictions, based on those two models. My workflow goes as follows: I start off with training the models. From both models I retrieve training predictions, which I get from cross_validation and the actual predictions on the test set. Afterwards, I fit a regressor with the training predictions and let the regressor make predictions on the actual predictions from the two models. The last part is where I'm stuck. What I'm doing now is simply stacking the predictions of the two models into a new array and feeding that array into my regressor. But that means I'm getting twice as many predictions/samples than I want. Do you have any idea on how I can combine these predictions in a way that I get the right amount of predictions/samples? |
