Incorrect processing of validation parameters
Created by: Dreamlone
For time series forecasting in-sample and cross-validation do not work correcty. Some cases I've checked:
-
'cv_folds': None, 'validation_blocks': None
- correct -
'cv_folds': None, 'validation_blocks': 1
- correct -
'cv_folds': None, 'validation_blocks': 2
- errors -
'cv_folds': 2, 'validation_blocks': None
- errors -
'cv_folds': 2, 'validation_blocks': 1
- correct -
'cv_folds': 2, 'validation_blocks': 2
- correct
In some cases even if marked as 'correct', Metric evaluation error arises It is required to correctly handle such combinations
Example for launch: https://github.com/nccr-itmo/FEDOT/blob/cv_folds/examples/simple/time_series_forecasting/api_foresacting.py
Estimated deadline - 17/06/2022