AffectLog

AffectLog: Trustworthy Machine Learning Assessment

Overview

Unverified black box models are destined for failure. Lack of transparency breeds distrust, leading to neglect and eventual rejection.

The 'affectlog' package offers a suite of tools to dissect and explain the behavior of any predictive model. The central component, the Explainer object, wraps around the model, facilitating detailed exploration and comparison through various model-level and prediction-level explanations. Additionally, 'affectlog' provides methods for assessing fairness and interactive dashboards for comprehensive analysis.

Installation

The affectlog package is available on PyPI and conda-forge.

pip install affectlog -U

conda install -c conda-forge affectlog

One can install optional dependencies for all additional features using pip install affectlog[full].

Resources: https://affectlog.com/research.html
API reference: https://affectlog.com/research/api

## Authors
The authors of the affectlog package are:

AffectLog Developer Team
We welcome contributions: start by opening an issue on GitHub.

## Citation
If you use affectlog, please cite our research:
@article{AffectLog360,
  author  = {AffectLog Developer Team},
  title   = {AffectLog: Trustworthy Machine Learning
             with Interactive Explainability and Fairness in Python},
  journal = {Research- AffectLog360°},
  year    = {n.d.},
  url     = {https://affectlog.com/research.html}
}

Changelog

v0.0.2 (2024-02-28)

• Dependencies:
  • Increased the dependencies to python>=3.8, pandas>=1.5.0, numpy>=1.23.3.
  • Added python==3.11 to CI.
• TensorFlow/Keras Compatibility:
  • Added keras.src.models.sequential.Sequential to classes with a known predict_function; this fixes changes in keras==3.0.0 and tensorflow==2.16.0.
  • Turned off verbose in the predict method of tensorflow/keras models to address changes in tensorflow>=2.9.0.
• Warnings and Errors:
  • Updated the warning occurring when specifying variable_splits.
  • Fixed an error occurring in predict_profile() when a DataFrame has MultiIndex in pandas>=1.3.0.
  • Fixed Gaussian norm() calculation in model_profile() from pi*sqrt(2) to sqrt(2*pi).
  • Fixed a warning (future error) between prepare_numerical_categorical() and prepare_x() with pandas==2.1.0.
  • Fixed a warning (future error) concerning the default value of numeric_only in pandas.DataFrame.corr() in affectlog.aspect.calculate_assoc_matrix().
• Explainer Enhancements:
  • Improved Explainer object to better handle new dependencies and compatibility issues.

v0.0.1 (2023-12-16)

• Precision and Recall Functions:
  • Added handling for ZeroDivisionError in precision and recall functions to prevent crashes.
• Warnings and Alerts:
  • Added a warning to calculate_depend_matrix() when there is a variable with only one value to notify users of potential issues.
• Exploratory Data Analysis (EDA) Plots:
  • Fixed missing EDA plots in the AL360 module, enhancing the visualization and analysis capabilities.
• Predict Parts Explanations:
  • Fixed baseline positions in the subplots of the predict parts explanations: BreakDown, Shap, ensuring accurate visual representation.
• Model and Predict Enhancements:
  • Improved model and predict functionalities to align with the latest updates and user feedback.