search

mljar / mljar-supervised

Python package for AutoML on Tabular Data with Feature Engineering, Hyper-Parameters Tuning, Explanations and Automatic Documentation
https://mljar.com
MIT License
3.02k stars 403 forks source link

bug for visualizing feature importance #427

Open akiliyiu opened 3 years ago

akiliyiu commented 3 years ago

image this is learner_fold_0_shap_summary.png saved from the decision tree trail, while the other .png saved in the same trail are all correct

pplonski commented 3 years ago

@akiliyiu thank you for reporting, could you please add CSV files that represent feature importance for that model? They have feature_importance in the file name and should be in the model folder. I will need them to reproduce and fix the bug.

akiliyiu commented 3 years ago

learner_fold_0_importance.zip

@akiliyiu thank you for reporting, could you please add CSV files that represent feature importance for that model? They have feature_importance in the file name and should be in the model folder. I will need them to reproduce and fix the bug.

pplonski commented 3 years ago

@akiliyiu thank you, looks like there might be a problem with features with float numbers in the name, like: hist_h_ptc_0.262_0.303. I will try to reproduce the issue locally and write unit test.