E-Mail Filtering Using Bayesian Network

E-Mail is important today. It is applied in many application; Education, Business and personal communication. Once there are too many E-Mail arrived in the mailbox and mostly are unwanted E-Mail, called Spam. Spam is a costly problem. At Prince of Songkhla University (PSU), there are around 5,000 e-mail users and around 40,000 messages received a day. There are 10% of them are virus and spam messages. Otherwise, the mail server has to pay memory and CPU load to process these virus and spam messages. These may cause the server response slowly and sometime once the system resources are insufficient, the mail server may crash and unavailable. Many filtering techniques are proposed. Bayesian Network is one of the popular Spam Filtering methods. This project is study Bayesian Network using SpamBayes, Open Source Software. Spam E-Mail are always written in English but at PSU there
are Thai Language Spam found increasingly. Thai Language is different from English Language because English word is separated by space but Thai Language is not. The project examines the SpamBayes accuracy on Spam classification of mix Thai and English E-Mail messages. Thai and English E-Mail are trained together and test messages are also Thai and English mixed. The result shows that SpamBayes
can classify Spam both in Thai or English.