01-31-2017 08:17 AM - edited 01-31-2017 08:37 AM
In a Miner project (EMiner 14.1), I've divided the input dataset in 4 bins and trained 4 different models. Now I'm trying to join them via an Ensemble node and to assess the performance, but when I do it the node takes only the test dataset of the first bin. I've also tried to append all the test datasets with an Append node and connect it to the Ensemble, but I cannot manage to use this "entire" test dataset.
Can you help me?
02-08-2017 10:24 AM
If the 4 models are of the same type (e.g. stepwise regression) then one could put the regression node between a start and end group node and model by segment, then add Model Comparison node.
If the 4 models are of different types, e.g. a tree, a regression, etc. then after appending the data with the append node, use a model import node before the model comparison node.
No Ensemble node required here.