Strongly consider implementing scikit-plot

[Aylr]

Benefits

• This could simplify and/or remove lots of code in feature importance, ROC, PR and confusion matrix plots
• This would mean less code for us to maintain.
• We would automatically gain new features, such as multiclass PR/ROC, learning rate curves, clustering graphs, and more
• No additional dependencies are introduced here (we already use matplotlib, scikit and scipy)

https://github.com/reiinakano/scikit-plot