Online Self-Organised Map Classifiers as Text Filters for Spam Email Detection

Email communication today is a way of working and communicating for most businesses and public in general. Being able to efficiently receive and send emails therefore becomes a must. Spam email detection and removal then becomes a vital process for the successful email communications, security and convenience. This paper describes a novel way of analysing and filtering incoming emails based on the text (keyword) salient features identified within. The method presented has promising results and at the same time significantly better performance than other statistical and probabilistic methods and at the same time offers a mechanism that can automatically adapt to new (unseen) email trends. The salient features of emails are selected automatically based on functions combining word frequency and other discriminating matrices, and then encoded into appropriate numerical vector models. The method is compared against the state-of-the-art Multinomial Naïve Bayes, Support Vector Machines and Boosted Decision Tress classifiers for identifying spam. The proposed automatic adaptable feature extractor method and online Self-Organising Map seems to give significantly better results, with the minimal cost.