Hi,
I have a complete set of data.
I only used 80% of it to train the two algorithms.
But, in principle, the best practice now is to determine which algorithm is the most efficient by testing them on the 20% of unused data, right?
Currently, I compared them after training on the 80%. But that's not best practice.
I suppose I need to use the second, unused branch of my Split Data block, which will correspond to my 20% of data.
But how do I express the fact that I want to use a model that has already been trained?
Because if I duplicate a new Linear Learner block, for example, as I understand it, this new block will not benefit from the training of the first block.
But perhaps I have misunderstood the meaning of this step?
Regards,
Denis
