OptBinning 0.20.0

Improvements:

• Axis labels are cut off when saving the table plot to file #303
• Packaging improvements #335

Bugfixes:

• Legend missing in ScorecardMonitoring.psi_plot() #327

OptBinning 0.19.0

Improvements:

• Adjust plot size #244.
• Save optimal binning object in JSON format #96.
• Plot IV/WoE metric in binning table plot for binary and continuous target.

Bugfixes:

• Keep pandas.DataFrame index in transform method #286.
• Fix BinningProcess's binning_transform_params="bins" #266 .

OptBinning 0.18.0

Bugfixes:

• Fix numpy array object #229
• Fix show_bin_labels #262
• Fix special_codes_y #263

OptBinning 0.17.3

Version 0.17.3 (2023-02-12)

Improvements:

• Implement sample_weight check in Scorecard class #228.

Bugfixes:

• Fix metric_missing ignored in Scorecard class #226.

Dependencies:

• Update RoPWR required version.

OptBinning 0.17.2

Improvements:

• Modify max-pvalue and min_diff constraints for CP and MIP formulation to avoid suboptimal solutions.

Bugfixes:

• Use keyword arguments in compute_class_weight #222.
• Remove preprocessing step when monotonic trend in (ascending, descending) for scenario-based binning #216 .

Dependencies:

• Update scikit-learn and ortools required versions.

OptBinning 0.17.1

New features

• Add parameter cat_unknown to assign values to the unobserved categories during training.

Improvements

• Add method decision_function to Scorecard #198.

OptBinning 0.17.0

New features:

• Optimize formulation of minimum difference constraints for all optimal binning classes and support these constraints regardless of the monotonic trend #201.

• Implementation of sample weight for ContinuousOptimalBinning #131.

Bugfixes:

• Fix ContinuousOptimalBinning prebinning step when no prebinning splits were generated #205.

OptBinning 0.16.1

New features:

• Outlier detectorYQuantileDetector for continuous target #203.

Improvements

• Add support to solver SCS and HIGHS for optimal piecewise binning classes.
• Unit testing outlier detector methods.

Bugfixes

• Pass lb and ub as keyword arguments to RoPWR fit method (required since ropwr>=0.4.0).

OptBinning 0.16.0

New features:

• Treatment of special codes separately for optimal piecewise binning classes #191.

Improvements

• Allow plot style="actual" for stochastic optimal binning.
• Unit testing optimal piecewise binning classes #93.
• Unit testing add macOS Monterey 12.

Bugfixes

• Fix sample weight for BinningProcess when n_jobs != 1 #190.
• Fix transform method for optimal binning 2D when dtype is categorical #197.
• Fix max_pvalue default value in documentation #199.

OptBinning 0.15.1

New features:

• New parameter show_bin_labels for binning tables #180.