Anomaly detection: do you apply transformation on new samples?

Course Q&A Machine Learning Specialization Unsupervised Learning, Recommenders, Reinforcement
1

Hello, we’ve seen on this section how to create new features / transform features on our training / cross-validation set to retrain the model and spot if the missing anomaly is now detected. When you move to productionising this model, would you apply that same transformation on the new sample before sending it to the model? My guess is yes, but would appreciate the confirmation.

Thanks

2

Hi! We do need to transform the samples for prediction like we transform the samples for training.

1 Like
3

Great, thanks for confirming :slight_smile: