K-means clustering scheme for enhanced spam detection

Abstract

In recent years, the problems of increasing spam mail on the internet are becomes a serious issue and difficult to detect. Furthermore, several e-mail classifications methods have been proposed and their performance is achieved. Although, Naïve Bayes classifiers (NB) has been widely used in e-mail classification and is very simple and efficient, yet the problem of improving the accuracy and reducing misclassification rate still exists. Therefore, many researches are being carried out. These studies propose a hybrid scheme for e-mail classification based on Naïve Bayes and K-means clustering to obtain better accuracy and reduce the misclassification rate of spam detection. The experiment of the proposed scheme was carried out using spam base benchmark dataset to evaluate the feasibility of the proposed method. The result of this hybrid led to enhance Naïve Bayes classifiers and subsequently increase the accuracy of spam detection and reducing the misclassification rate. In addition, experimental results on spam base datasets show that the enhanced Naïve Bayes (KNavie) significantly outperforms Naïve Bayes and many other recent spam detection methods. © Maxwell Scientific Organization, 2014.