A New Approch to Design of an Email Classifier

Due to the increase in the number of spam emails recently, the need of anti-spam filters has increased. To filter the spam emails we generally use Machine Learning Technique now a days. Naive Bayes is an open source anti-spam email filters which is widely used. In this algorithm, probability of the words occurring in the ham and spam mails is calculated during the training phase. Based on these trained set a message is tested for spam or legitimate mails. In our work, we have designed two other algorithms whose initial part are same as that of the Naive Bayes algorithm, but these two algorithms do not use the probability measure. Instead we have derived the Effective Spam words and Effective Ham words from training set and got two set of trained word list-one consisting of spam words and other consisting of ham words. Based on these two sets the testing of mails is done. The test result is presented in the project.