List your company
ResearchFormerly WizWand

Local Feature Importance Evaluation on 12 Real-World Datasets Aggregate (test)

1Average RankLMDI+
0.88681.65092.4153.1791Jun 10, 2025
Updated 5mo ago

Evaluation results

MethodLinks
LMDI+Top Features Retained=30%, Evaluation Protocol=Remove-and-Retrain, Base Model=Random Forest, Averaging=20 runs per dataset2025.061
LMDI+Top Features Retained=40%, Evaluation Protocol=Remove-and-Retrain, Base Model=Random Forest, Averaging=20 runs per dataset2025.061.08
LMDI+Top Features Retained=20%, Evaluation Protocol=Remove-and-Retrain, Base Model=Random Forest, Averaging=20 runs per dataset2025.061.17
LMDI+Top Features Retained=10%, Evaluation Protocol=Remove-and-Retrain, Base Model=Random Forest, Averaging=20 runs per dataset2025.061.25
LIMETop Features Retained=10%, Evaluation Protocol=Remove-and-Retrain, Base Model=Random Forest, Averaging=20 runs per dataset2025.062.17
LIMETop Features Retained=20%, Evaluation Protocol=Remove-and-Retrain, Base Model=Random Forest, Averaging=20 runs per dataset2025.062.17
LIMETop Features Retained=40%, Evaluation Protocol=Remove-and-Retrain, Base Model=Random Forest, Averaging=20 runs per dataset2025.062.33
LIMETop Features Retained=30%, Evaluation Protocol=Remove-and-Retrain, Base Model=Random Forest, Averaging=20 runs per dataset2025.062.42
TreeSHAPTop Features Retained=10%, Evaluation Protocol=Remove-and-Retrain, Base Model=Random Forest, Averaging=20 runs per dataset2025.062.83
TreeSHAPTop Features Retained=20%, Evaluation Protocol=Remove-and-Retrain, Base Model=Random Forest, Averaging=20 runs per dataset2025.062.83
TreeSHAPTop Features Retained=30%, Evaluation Protocol=Remove-and-Retrain, Base Model=Random Forest, Averaging=20 runs per dataset2025.062.83
TreeSHAPTop Features Retained=40%, Evaluation Protocol=Remove-and-Retrain, Base Model=Random Forest, Averaging=20 runs per dataset2025.063.08
Local MDITop Features Retained=40%, Evaluation Protocol=Remove-and-Retrain, Base Model=Random Forest, Averaging=20 runs per dataset2025.063.5
Local MDITop Features Retained=10%, Evaluation Protocol=Remove-and-Retrain, Base Model=Random Forest, Averaging=20 runs per dataset2025.063.75
Local MDITop Features Retained=30%, Evaluation Protocol=Remove-and-Retrain, Base Model=Random Forest, Averaging=20 runs per dataset2025.063.75
Local MDITop Features Retained=20%, Evaluation Protocol=Remove-and-Retrain, Base Model=Random Forest, Averaging=20 runs per dataset2025.063.83