Pattern Recognition with Nearest Neighbors

[bbengfort]

This post is the first in a series that discusses the machine learning methods that I will present in the Georgetown data analytics course. My hope is that I can expose through some tricky JavaScript/visualization the mechanism behind the learning as well as provide some description. These posts will go on DDL's blog.

For kNN the idea is to provide a visualization, that allows you to slide the k parameter to change the model. See Understanding the Bias Variance Trade-Off for something very similar. I will also provide accuracy metrics, as well as possibly a ROC curve or something similar. This will hopefully better allow students to understand how kNN works in practice.