Email Representation using Noncharacteristic Information and its Application

Abstract

Focusing on the uncertainty of classifying emails based-on email content and the incompleteness of email representation, the paper proposes a new representation using noncharacteristic information. The new approach refers to the whole email, contains feature items extracted from email content, and noncharacteristic items extracted from email header. In the expriment, we adopt Naive Bayes classifier to classify emails, classification results indicate that the new approach overcomes the shortcomings of original content-based filtering and improves the recall and the precision of spam filtering.