A novel approach for Spam Email Filtering Using Machine Learning

Abstract

Spam emails also known as unsolicited emails (maybe commercial or maybe not) i.e. those mails which are sent without our request or concern. Email spam is the practice of sending unwanted emails, mostly contains commercial messages to randomly generated persons. In the internet email spam is widespread because of such low cost of sending emails other than any other means of communication. It is important to filter spam emails because most of the malicious activities performed in the internet done through email spamming. Though there are many spam filters are available we still get huge amount of spam emails. This is not because the filters are not accurate ; effective; the reason is the generation of quick and effective counters of the algorithm used in the filters. In our project we used mainly three supervised learning algorithms namely Linear SVC, Multinomial NB, and k-NN to implement the filter. We used these algorithms to train the system about spam email by using the feature called word count vector which is generated by processing a dataset filled with existing emails containing both spam and legitimate emails. The full process of the project and the result of the execution by implementing the three models/algorithms are discussed.