psamknn: k-Nearest Neighbor classification on the PSAM spam database

Bart Massey

This repository contains a couple of things:

• Instances (feature-vectors) from the PSAM spam corpus of some years ago. Each csv file contains instances consisting of a name, a class (1 for spam, 0 for ham), and a vector of features obtained via big-bag-of-words and SpamAssassin analyses.

• Python code for a k-Nearest Neighbor classifier for the instances.

Running

You will need the Python gmpy2 package to run this. On Debian, you can say:

apt install libmpfr-dev
apt install libmpc-dev
pip3 install gmpy2

To run on the "personal" corpus with 10-way cross-validation and a neighbor distance of 5, say

python3 knn.py personal.csv 10 5

The output will consist of the accuracy for each cross-validation split.