pavelkrizek commented 3 years ago

Prevent situation when model incompatible data are not seen during CV (in the last split of actuals) and model which cannot be fitted on the full dataset is chosen. In such a situation the next model with the second-best error should be choosen.

codecov-io commented 3 years ago

Codecov Report

Merging #60 (4f54f6b) into master (33fcf2e) will increase coverage by 0.08%. The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master      #60      +/-   ##
==========================================
+ Coverage   93.61%   93.70%   +0.08%     
==========================================
  Files          56       56              
  Lines        2633     2670      +37     
==========================================
+ Hits         2465     2502      +37     
  Misses        168      168

Impacted Files	Coverage Δ
src/hcrystalball/model_selection/__init__.py	`100.00% <100.00%> (ø)`
...l/model_selection/_large_scale_cross_validation.py	`91.57% <100.00%> (+0.99%)`	:arrow_up:
...stalball/model_selection/_model_selector_result.py	`82.85% <100.00%> (+0.24%)`	:arrow_up:
tests/unit/model_selection/test_model_selection.py	`100.00% <100.00%> (ø)`

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 57e90b9...4f54f6b. Read the comment docs.

MichalChromcak commented 3 years ago

@pavelkrizek Thanks. Merging now 👍 Note from the review - changes should be backward compatible

heidelbergcement / hcrystalball

Fix problem with unstable models which are picked during CV #60

Codecov Report