Yifan Liu Undergraduate Dissertation 2014/15

Design Automatic Spam Filter based on Machine learning approach

Supervised by E.Vasilaki

Abstract

Nowadays, due to speedy transmission and exceedingly low cost, E-mail has become one of the most powerful communicate tool in Internet. However, with rapidly grow usage of E-mail, more and more unsolicited bulk e-mails such as ‘Amazing health tonics’, ’ Unleash your carnal needs’ and ’Invoice Attached’ are sent by social networks, swindler and advertisers. These spams cause a lot of trouble to the E-mail users.

Even a lot of Spam filter has been built for E-mail spam classification. None of these filters reach perfect accuracy in filtering spam E-mail. In this project, Backpropagation, Support Vector Machine and EM with GMM are used in analyse UCI spam dataset. Comparing and evaluating performance of each algorithm explore the efficient classification algorithm.