A Case-Based Technique for Tracking Concept Drift in Spam Filtering

Abstract

Clearly, machine learning techiques can play an important role in filtering spam email because ample training data is available to build a robust classifier. However, spam filtering is a particularly challenging task as the data distribution and concept being learned changeds over time. This is a particularly awkward form of concept drift as the change is driven by spammers wishing to circumvent the spam filters. In this paper we show that lazy learning techniques are appropriate for such dynamically changing contexts. We present a case-based system for spam filtering called ECUE that can learn dynamically. We evaluate its performance as the case-base is updated with new cases. We also explore the benefit of periodically redoing the feature selection process to bring new features into play. Our evaluation shows that these two levels of model update are effective in tracking concept drift.