Performance Evaluation of Two e-mail Filtering Systems with Different Machine Learning Techniques

We have tried to compare the accuracy rates of classification methods in two spam mail filter systems with different machine learning techniques. One of these systems employs BONSAI and the other employs support vector machine (in short SVM) as a machine learning techniques. These two systems can classify e-mails into the sets of spam mail or non-spam mail according to the appearing rate of words in combination with the sequence analysis based on its appearing rate. The mail filtering system using BONSAI could classify e-mails into the sets of spam mail or non-spam mail with more than 90% accuracy. The other system using SVM could also classify e-mails with more than 90% accuracy. Although two systems could classify e-mails into two categories with high accuracy rate, the system using BOSAI was characterized by an efficient classification on the condition that the less learning examples were given.