Stacking ML Algorithms

Question

Hello, I am new to ML and stacking and I still learning. I want to design an experiment in Azure ML Studio (classic) where I use stacking to combine PCA-based anomaly detection, one-class SVM, and two-class neural network. Can I stack these three methods where I use only the Tune Model Hyperparameter instead of combining Tune Model Parameter and Train Anomaly Detection for PCA-based and one-class SVM-based anomaly detection and Train Model for the two-class neural network? Also, to train the PCA-based anomaly detection as a meta model, can I use the Train Model Hyperparameter in conjunction with the Train Anomaly detection method? For the base models can I hyper tune parameters and then use the appropriate train model?

Answer 1

Hello,

You can use different algorithm to train your model if we don't consider too much about the effectivity. But one thing to mention here is, you need to suit your output date from model 1 to model 2 well as input data.

Regards,
Yutong

Answer 2

Thank you! That was helpful.