Ensemble Learning: Stacking

Stacking Classifiers in

Dataset

The evaluation is conducted over a two-dimensional dataset consisting of 1200 observations.

Visual representation of the standardized dataset.

As shown in , the data features a non-linear, circular distribution where one class is nested within another.

This geometry suggests that linear classifiers may struggle compared to kernel-based or neighborhood-based models.

Models

Results and Discussion

Model	Accuracy
Nearest Neighbors	0.973
QDA	0.990
RBF SVM	0.980
Decision Tree	0.952
Random Forest	0.957
Gradient Boosting	0.972
AdaBoost	0.978
Neural Network	0.985
Hard Meta-Classifier	0.982
Soft Meta-Classifier	0.995

Mean classification accuracy scores for different classifiers...

	Standard CV	Data Leakage (Prefit)
Hard Meta-Classifier	0.982	0.978
Soft Meta-Classifier	0.995	0.972

Mean classification accuracy scores for different classifiers...