I split my dataset into train and test data. Then I trained logistic regression, SVM, and random forest models using pipelines with cross-validation and train data.
I saved best-performing models with tuned parameters via best_estimator_
And now I want to make a voting ensemble with these models to predict using test data.
The questions are:
- Can pre-trained models be used in ensembles this way?
- On which data I should train the ensemble?