yellowbrick

Why we included this project

Yellowbrick is for the part of ML work that raw metrics don't show. Instead of staring at a single accuracy score, you get visualizers that draw ROC curves, confusion matrices, learning curves, and feature rankings straight from your own estimator and data. Each visualizer is a thin wrapper around a matplotlib figure that plugs into the normal scikit-learn fit workflow, so in a notebook it is a few lines to drop in and the results stay reproducible. It also covers feature analysis and hyperparameter tuning, which is where you find out why a model underperforms before you start tuning blindly. It is a companion to scikit-learn rather than a replacement, aimed at people who want to see the model instead of trusting a number.