Assignment 7: Naive Bayes

Assignment Description

In this assignment, you will implement a spam detector for text messages using Naive Bayes. The natural-language-processing aspects of our implementation are much simpler than what you should do in practice, but they will suffice to illustrate the power of Naive Bayes. The main advantage of Naive Bayes is its simplicity, together with its relative accuracy (given how incredibly straightforward it is, it’s surprising that it works at all!)


Answer the questions below in a “answers.txt” plain file, “” Markdown, or “answers.pdf” PDF. I will not accept Microsoft Word, OS X Pages, or OpenOffice documents. (I prefer Markdown, so I can see it from your repository on Github directly)

In addition, submit whatever code you use to answer the questions below.

Naive Bayes hyperparameters

In your code, you will have to choose a “smoothing” hyperparameter that will be used as a pseudocount. Experiment with a variety of different values.

Implementation details

If you follow the first formulation in the book (page 122, equation 9.25), then your implementation will be quite inefficient. Why is that, and how can you use Equations 9.27 and 9.28 to improve it?


The dataset for this assignment comes from UCI, courtesy of Tiago A. Almeida and Jose Maria Gomez Hidalgo.

(If you end up inspecting the data, do note that this is a real-world dataset, and comes with a certain amount of adult language.)