-
The architecture of a stacking model involves two or more base models, often referred to as level-0 models, and a meta-model that combines the predictions of the base models, referred to as a level-1 model.
-
We have used the following models as level-0 models:
- Gaussian Naive Bayes Classifier (With hyperparameter tuning)
- Random Forest Classifier (With hyperparameter tuning & 3-fold CV)
- Decision Tree Classifier (With hyperparameter tuning & 3-fold CV)
- SVM Classifier (With hyperparameter tuning & 3-fold CV)
- ANN Model (With hyperparameter tuning & 3-fold CV)
- Logisitic Regression (With hyperparameter tuning & 3-fold CV)
-
We have used simple "Logistic Regression" Model (from python's 'scikit' module) as the level-1 model, with 4-fold CV.
-
We have achieved 82.68% test accuracy on "Our Model", which is better than all the 6 individual models.
Vinay Khilwani, Vasu Gondaliya, Shreya Patel, Jay Hemnani & Bhuvan Gandhi