how to divide data into data testing and training
1 view (last 30 days)
Show older comments
hai everyone i want to ask about my problem i want to predicting exchange rate using anfis and i confuse how to divide data into data testing and training ?
2 Comments
José-Luis
on 7 Jul 2017
Edited: José-Luis
on 7 Jul 2017
Depends on your objective:
- If you're a cynic only interested in results: train with as much as possible and test with the rest
- If you're a highly-principled masochist interested in the robustness of your model: use as little as possible for training and test with the rest.
Disclaimer: Jokes happen
Accepted Answer
Greg Heath
on 8 Jul 2017
The NN Toolbox default is
15% Testing
85% Design
--- 70% Training
--- 15% Validation
The validation subset is a non-training design subset used during training to make sure that the currently trained net generalizes well to non-training data.
The current default setting of 6 insures that training will stop if the non-training validation subset error increases continually for 6 epochs.
Sample code from the documentation obtained via the commands help and doc uses all defaults. For regression/function-fitting/curve-fitting the default function is FITNET. The sample code using as many defaults as possible is obtained via the commands
help fitnet
and
doc fitnet
Alternative code by me is obtained from
Hope this helps
Thank you for formally accepting my answer
Greg
0 Comments
More Answers (0)
See Also
Categories
Find more on Sequence and Numeric Feature Data Workflows in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!