Spam mail detection using classification techniques and global training set

Abstract

Emails are Internet-based services for various purposes like sharing of data, sending notices, memos, and sharing data. Spam mail are emails that are sent in bulk to a large number of people simultaneously, while this can be useful for sending same data to a large number of people for useful purposes, but it is mostly used for advertising or scam. These spam mails are expensive for the companies and use a huge amount of resources. They are also inconvenient to the user as spam uses a lot of inbox space and makes it difficult to find useful and important emails when needed. To counter this problem, many solutions have come into effect, but the spammers are way ahead to find these solutions. This paper aims at discussing these solutions and identifies the strengths and shortcomings. It also covers a solution to these spam emails by combining classification techniques with knowledge engineering to get better spam filtering. It discusses classification techniques like Naïve Bayes, SVM, k-NN, and Artificial Neural Network and their respective dependencies on the training set. In the end of this paper, the global training set is mentioned which is a way to optimize these training sets and an algorithm has been proposed for the same.