The annoyance of spam increasingly plagues both individuals and organizations. Spam classification is an important issue to distinguish the spam with the legitimate email or address. This paper presents a neural network ensemble approach based on a specially designed cooperative coevolution paradigm. Each component network corresponds to a separate subpopulation and all subpopulations are evolved simultaneously. The ensemble performance and the Q-statistic diversity measure are adopted as the objectives, and the component networks are evaluated by using the multi-objective Pareto optimality measure. Experimental results illustrate that the proposed algorithm outperforms the traditional ensemble methods on the spam classification problems.
Tian, Jin; Li, Minqiang; and Chen, Fuzan, "An Effective Ensemble Approach for Spam Classification" (2010). ICEB 2010 Proceedings. 25.