Automodel - Performance
Maven
I have 1.281 lines of data on my set. I launched 2 models in Auto-Model feature.
Both have based their performance on 512 lines, which is +/-40% of the lines, as Training Data.
Is this normal ratio? And how should I use the remaining 60% of lines to Validate my Data? After deployment?
Many thanks for your feedback.
Both have based their performance on 512 lines, which is +/-40% of the lines, as Training Data.
Is this normal ratio? And how should I use the remaining 60% of lines to Validate my Data? After deployment?
Many thanks for your feedback.
Tagged:
0
Best Answer
-
lionelderkrikor
Moderator, RapidMiner Certified Analyst, Member Posts: 1,195 
Unicorn
Hi @Olus,
I think you are doing a little misunderstanding :
To sum up, Auto-Model is using :
- 60% of your initial dataset to train the model(s) : It's the training set
- the remaining 40% of your initial dataset to test the model(s) : It's the test set.
To be more accuracy, here an extract of the documentation of how is calculated the performance of a model in Auto-Model.
You can find this documentation by clicking in the "I" (at the top right of the screen) in the final screen of Auto-Model (results screen) :
Regards,
Lionel
6
Answers
Many thanks for your precious time and help on answering my question.
Regards,
Olus