I read a paper recently where researchers were trying to do classification using ensembling methods. I first read about the concept of BOOSTED-STACKING there.For those who dont know , I can give a brief overview.There are mainly 4 types of ensembling techniques:
This link will give you a quick insight into the topic. https://stats.stackexchange.com/questions/18891/bagging-boosting-and-stacking-in-machine-learning
So after reading about BOOSTED-STACKING , I came across a pretty neat algorithm for BOOSTED-STACKING.
Step 1: Import the dataset Step 2: Give 75% data for training and remaining data for testing. Step 3: Apply “n” different base learners for AdaBoost Step 4: Assign train dataset to the models Step 5: Update the weight based on the misclassification rate Step 6: Choose the best learners among “n” different base learners Step 7: Apply the selected learners for AdaBoost Step 8: Now create the best combination for AdaBoost Step 9: Create the model using Stacking. Step 10: Make predictions for test dataset and calculate accuracy.
I think I understood this. But then I thought what imrovements can be done here? Also How will they implement other hybrid ensembles. So I read some more and found this link https://medium.com/syncedreview/infiniteboosting-a-bagging-boosting-hybrid-algorithm-8b109019e480 which explains about Bagging Boosting Hybrid. And I understood that also.
Now I am wondering how can we implement other combinations of ensemble methods that I thought would be cool like maybe BOOSTED-BLENDING etc.
I am still a beginner in this domain and I do not have the mindset yet to come up with these methods on my own and be confident about them. So to summarise it all, can you please explain to me the pseudo-algorithms for these ensemble methods like Bagged-Stacking and Boosted-Blending etc (if it is even logical and possible to implement these hybrids).