Tweaking naive Bayes classifier for intelligent spam detection
Ankita Raturi, Sunil Pranit Lal · Massey Research Online (Massey University) · 2011
Spam classification is a text classification task that is commonly implemented using Bayesian learning.These classification methods are often modified in order to improve the accuracy and minimize false positives.This paper describes a Naïve Bayes (NB) classifier for basic spam classification.This is then augmented with a cascaded filter that uses a Weighted-Radial Bias Function (W-RBF) for similarity measure.It is expected that the NB classifier will perform the basic classification with the W-RBF acting as a secondary filter, thus improving the performance of the spam classifier.It was found that the NB portion of the cascade was the initial spam filter with the W-RBF filter acting as a False Positive filter.