0.0.8

Version 0.0.8 🚀

• Removing quantiles parameter, so it will be only necessary to specify alpha;
• Added Winkler Interval Score and MSE in Evaluate callable function in Conformal Regressors

0.0.7

Version 0.0.7 🚀

📚 Improved Documentation:

• Enhanced function documentation for better clarity and understanding.
• Removed outdated references and mentions that are no longer relevant to the current implementation.

🐛 Coverage Rate Fix:

• Simplified the coverage_rate method to directly check for coverage in the selected data.
• This change improves reliability and makes it easier to verify coverage with new datasets.
• ⚠️ Breaking Change: Replaced _empirical_coverage with this method.

🐛 ClassConditional Adjustment:

• Fixed an issue where q_level was not being properly assigned to the volume of the specified class group in the ClassConditional classifier.

✨ New Regression Models:

• Added support for Conformalized Quantile Regression (CQR) and Conformalized Regression models.
• These models provide robust prediction intervals for regression tasks, expanding the library's capabilities.

0.0.6

Version 0.0.6 🚀

Latest Release

🛠️ Minor Updates

• ✨ Add False Positive Rate in evaluate function:
  Now includes FPR in evaluation results for better significance level assessment.

• 🔥 Remove evaluate generalization:
  This metric was removed due to inconsistent results caused by test/train distribution mismatches.

• 🐛 Fix evaluate predict call:
  Resolved an issue where predictions weren't properly adjusting based on the alpha parameter.

0.0.5

Version 0.0.5 🚀

Latest Release

🛠️ Main Updates

• ♻️ Add total in evaluate:
  Added the total count in the evaluation metrics.

• 🐛 Random seed in shuffle:
  Fixed an issue with the random seed in the shuffle method.

• ♻️ Add option to fit Conformal whether OOB or not:
  Introduced an option to enable/disable OOB (Out-Of-Bag) fitting in the Conformal classifier.

• ♻️ Change class names:
  Renamed classes to remove the OOB prefix, e.g., OOBBinaryClassConditionalConformalClassifier to BinaryClassConditionalConformalClassifier.

0.0.4

Version 0.0.4 🚀

Latest Release

🛠️ Main Updates

• ✅ Add unit tests: unit tests have been added to ensure code reliability and correctness.

• 🚚 Moving Examples from a directory specific: Examples have been reorganized from a specific directory to a more generalized location.

• 🐛 Modularize functions from OOB classes for better organization: Functions within OOB (Out-Of-Bag) classes have been modularized to enhance code readability and maintainability.

• ♻️ Generalization Score by Balanced Score Accuracy: The generalization score calculation has been updated to use Balanced Score Accuracy, providing a more reliable evaluation metric for the classifier's performance across different datasets.

0.0.3

Version 0.0.3 🚀

Latest Release

🛠️ Main Updates

• 🚚 Reorganized Examples:
  Moved all example files to a dedicated folder for better organization.

• 🐛 TOML Fixes:
  Adjusted the toml configuration. The installation of plotting tools is no longer required.

• 🐛 Generalization Score Fix:
  Normalized the score to correctly reflect the difference between two values.

• ♻️ Evaluate returning dict:
  Updated the evaluate function to return a dictionary instead of an unnecessary DataFrame.

0.0.2

Version 0.0.2 🚀

Latest Release

🛠️ Key Changes

• ♻️ Improved Calibration:
  In addition to using the Balanced Accuracy Score, calibration can now be performed with:

  • Matthews Correlation Coefficient
  • Bookmaker Informedness
    These options provide enhanced reliability for evaluation.

• ♻️ Class Renaming:
  The BaseConformalClassifier class has been renamed to BaseOOBConformalClassifier to better reflect its functionality and usage.

0.0.1

Version 0.0.1 🚀

What's New

This is the first release of my project! 🎉 Here are the main changes and features included:

Main Features

• 🌟 OOBBinaryClassConditionalConformalClassifier: A modrian class conditional conformal classifier based on Out-of-Bag (OOB) methodology, utilizing a random forest classifier as the underlying learner and Venn-Abers calibration..
• 🌟 OOBBinaryMarginalConformalClassifier: Conformal classifier based on Out-of-Bag (OOB) predictions. based on Out-of-Bag (OOB) methodology, utilizing a random forest classifier as the underlying learner and Venn-Abers calibration.

---

Release Date: January 9, 2025