This release add following new features:
Feature engineering
- Feature generation
- Feature selection
Data clean
- Special empty value handing
- Correct data type
- Id-ness features cleanup
- Duplicate features cleanup
- Empty label rows cleanup
- Illegal values replacement
- Constant features cleanup
- Collinearity features cleanup
Data set split
- Adversarial validation
Modeling algorithms
- XGBoost
- Catboost
- LightGBM
- HistGridientBoosting
Training
- Task inference
- Command-line tools
Evaluation strategies:
- Cross-validation
- Train-Validation-Holdout
Search strategies
- Monte Carlo Tree Search
- Evolution
- Random search
Imbalance data
- Class Weight
- Under-Samping
- Near miss
- Tomeks links
- Random
- Over-Samping
- SMOTE
- ADASYN
- Random
Early stopping strategies
- max_no_improvement_trials
- time_limit
- expected_reward
Advance features:
- Two stage search
- Pseudo label
- Feature selection
- Concept drift handling
- Ensemble