pmaji / data-science-toolkit Goto Github PK

Need to swap out the default colors for a chosen palette for originality. Explore use of ggthemes as well in R code.

Had a great discussion with @TWilliamBell as a result of my reddit post sharing the work I had started on this rep.

I believe the fruit of that conversation will be the addition of some materials under the classification section that will cover KNN methods building off of / summarizing the excellent work already started over in this repo here.

Opening this issue for the discussion of all things pertaining to the prospective KNN section, and I have opened a subfolder for KNN under classification with a placeholder text file for now.

Work to do in the future: replicate the vignette I constructed for logistic regression in R but using Python instead.

Resources to draw from include @nyilmaz1's repository linked below (found via a discussion we had on LinkedIn).

https://github.com/nyilmaz1/Breast-Cancer-predictions/blob/master/Breast%20cancer%20predictions-WITH%20LIME%20AND%20VALIDATION.ipynb

This a couple of items away on my priority list in terms of projects of this type, but @ericctbai expressed interest and getting starting on this work.

@ndrewGele is going to put together a PR that involves a functionalized solution for density plot creation to supplement the md on the KS test here:

https://github.com/pmaji/data-science-toolkit/blob/master/hypothesis-tests/ks_test.md

Need to include some additional commentary in the markdown--along with any necessary or pertinent tests--that discuss assumptions necessary when building logits.

When I scale up the KS test to be performed multiple times, I should look into applying some sort of a correction to account for the multitude of trials. Right now, I am leaning towards the simple but effective Bonferonni Correction, linked below.

https://en.wikipedia.org/wiki/Bonferroni_correction