Building the stacking classifier
Now you'll work on the next two steps.
Step 3: Append the predictions to the dataset:
this is internally handled by the StackingClassifier
class, but we'll do our part by preparing the list of first-level classifiers, which you built in the previous exercise. These are available as: clf_dt
and clf_knn
.
Step 4: Build the second-layer meta estimator:
for this purpose you'll use the default LogisticRegression
. This will take as input features the individual predictions from the base estimators.
With both levels of estimators ready you can build the stacking classifier.
This exercise is part of the course
Ensemble Methods in Python
Exercise instructions
- Prepare the list of tuples with the first-layer classifiers:
clf_dt
andclf_knn
(specifying the appropriate labels and order). - Instantiate the second-layer meta estimator: a
LogisticRegression
. - Build the stacking classifier passing: the list of tuples, the meta classifier, with
stack_method='predict_proba'
(to use class probabilities), andpassthrough = False
(to only use predictions as features).
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
# Prepare the list of tuples with the first-layer classifiers
classifiers = [
____,
____
]
# Instantiate the second-layer meta estimator
clf_meta = ____
# Build the stacking classifier
clf_stack = ____(
____,
____,
____,
____)