Unit:
Τομέας Υπολογιστικών Συστημάτων και ΕφαρμογώνLibrary of the School of Science
Author:
Γιαννόπουλος Αθανάσιος
Supervisors info:
Παναγιώτης Σταματόπουλος
Original Title:
Αναγνώριση ανεπιθύμητης αλληλογραφίας με χρήση νευρωνικών δικτύων
Translated title:
Spam filtering using neural networks
Summary:
In the period that Spam emails deluge every mailbox available, the need of
automated
recognition and filtering seems mandatory. During this project, the techniques
of Email
Spam Filtering used in the latest years were studied. Specifically, more
attention was given
to the approaches that use Machine Learning Algorithms. Subsequently, a Spam
Filter
was developed using Neural Networks and more specifically the Multilayer
Perceptron
/ MLP algorithm. The collection of Machine Learing and Data Mining algorithms
Weka
developed in University of Waikato was used. The Neural Network’s parameters
were
studied and the optimal values are decided in order to maximize classification
accuracy
of the Spam Filter. Both the techniques that were used and the experimental
results are
reported in this project.
Keywords:
email, spam filtering, machine learning, neural networks, tf-idf
