Spam Filter Based on Multiple Classifiers Combinational Model

To overcome the shortages of the limited application scope and the low accuracy of single classifier at spam filtering, a combinational model, which is based on naive Bayesian, k-nearest neighbor (KNN), and support vector machine (SVM), is proposed in this chapter. The performance is improved in this model by using the voting rule and human-computer interactions. The comparison of anti-spam filters based on the four models, respectively, with experiments on public mail corpora is described. The experiments show that Multiple Classifier Combinational Model (MCCM) achieves better performances than the best single one. With the analysis of the different flitters, it is concluded that the MCCM is more suitable for anti-spam filtering. ? Springer-Verlag Berlin Heidelberg 2014.