PR: adding comparisons between models

[csbell-vu]

Summary

• Adds comparative ROC curves and calibration plots to the 55-cv-comparison notebook
• Final fits and other parameters for the generated models are stored on Box
• Training html, performance html, and comparative html results in html_results folder

Findings summary

• The random forest ROC curve is extremely suspicious
• All models are well calibrated (and scores can reasonably be interpreted as probabilities) except for Naive Bayes

[csbell-vu]

A few more commits were added to this branch to do the following:

• Simplify the load process using Box Sync (the default file path is generally relative to ~/Documents than ~ so this was explicitly added to the Box filepath)
• Permutation importance plots were updated to use permutation importance via workflows rather than the underlying model fit. This assisted in removing ambiguity about what class was being used for prediction and metric assessment.
• Updated all html outputs for notebooks with updated permutation importance plots, and added additional commentary on plots for better interpretation