familiar: familiar: Fully Automated Machine Learning with Interpretable...

In familiar: End-to-End Automated Machine Learning and Model Evaluation

familiar: Fully Automated Machine Learning with Interpretable Analysis of Results

Description

End-to-end, automated machine learning package for creating trustworthy and interpretable machine learning models. Familiar supports modelling of regression, categorical and time-to-event (survival) outcomes. Models created using familiar are self-containing, and their use does not require additional information such as baseline survival, feature clustering, or feature transformation and normalisation parameters. In addition, an novelty or out-of-distribution detector is trained simultaneously and contained with every model. Model performance, calibration, risk group stratification, (permutation) variable importance, individual conditional expectation, partial dependence, and more, are assessed automatically as part of the evaluation process and exported in tabular format and plotted, and may also be computed manually using export and plot functions. Where possible, metrics and values obtained during the evaluation process come with confidence intervals.

Author(s)

Maintainer: Alex Zwanenburg [e-mail] (ORCID)

Authors:

• Steffen Löck

Other contributors:

• Stefan Leger [contributor]

• Iram Shahzadi [contributor]

• Asier Rabasco Meneghetti [contributor]

• Sebastian Starke [contributor]

• Technische Universität Dresden [copyright holder]

• German Cancer Research Center (DKFZ) [copyright holder]

See Also

Useful links:

• https://github.com/alexzwanenburg/familiar

• Report bugs at https://github.com/alexzwanenburg/familiar/issues

familiar documentation built on Sept. 30, 2024, 9:18 a.m.