An Empirical Study of the Naive Bayes Classifier

The Paper:
Naive Bayes classifiers are based upon Bayes theorem, a probability and stats theory that relates current probability to prior probability. A central assumption of the Naive Bayes classifier is the assumption that each predictor, or piece of evidence observed, is independent of other predictors. Given this simplicity (the 'naive' assumption) Naive Bayes classifiers have been shown to outperform other more sophisticated classifiers. We'll review what types of data attributes Rish argues can impact the performance of Naive Bayes as well as review an implementation in Python I wrote.

http://www.cc.gatech.edu/~isbell/reading/papers/Rish.pdf

The Speaker:
Given that Lorena was a policy analyst what is the likelihood that yes she would want to learn about modeling data? (The overwhelming likelihood says yes - she does.) During the day Lorena works as a software engineer at Sprout Social and by night, she likes to read papers like this one (as well as play Mortal Kombat and/or watch way too much Star Trek). Oh, and if you're a data enthusiast too, tell Lorena you want to try out a Kaggle challenge with her. She is really into the idea of predicting altruism through free pizza ( https://www.kaggle.com/c/random-acts-of-pizza ).