adding KNN to cl_bayes and modularizing

[sophiacofone]

Hello,

I was looking more closely at Connor's cl_bayes file and I think that the code there would work really well with KNN model. My understanding of KNN leads me to think that it would sort-of be grouped in with these other "decision boundary"/distance-type models. I was wondering if you guys agree that KNN would be good to add?

Secondly, I quickly tested adding KNN classifier to the main function and it seems like it works, but would need some tweaking to get the plotting working (for example adding additional subplot). Before I tweak more things, I was wondering if it might make sense to "modularize" the script to be able to incorporate other features. My impression is that right now is that it is focused on ['boxratio', 'boxsize/interoc'] features, but could probably be changed to incorporate other "columns" (like Jesse and I were doing with logreg).

Any thoughts on these items?

[CCLynch]

Closing this issue. KNN has been added to cl_bayes.py. @sophiacofone raised good points in #25 that the models could be explored further with euclidean distances dataset and the feature selection she created with sc_feature_selection.py I'll add another issue for that.