Polynomial neural networks versus other spam email filters: An empirical study

Abstract

Spam or junk e-mail problems are increasing exponentially due to the huge growth of internet users and their high dependency on e-mails as the main communication means nowadays. Such problems result in huge amounts of time and cost waste for both individuals and organizations. This research paper directly compares the performance of four famous text classification algorithms in classifying emails and detecting the spam ones: Polynomial Neural Networks (PNNs), the k-nearest neighbour (k-NN), Support Vector Machines (SVM) and Naïve Bayes (NB). Results of the experiments conducted on Lingspam, the benchmark E-mail corpus, in this research work reveals that PNNs is a competitive spam filter to the sate-of-the-art spam filters. It recorded either equal or superior results in most of the performance measures used to evaluate the four spam filters.