Email Spam Classifications Based on Support Vector Machine and Recurrent Neural Network

Satyendra Shukla; Pawanesh Singh

Abstract

In recent times, e-mail has become one of the fastest and the utmost economical process of communication. Spam emails have dramatically increased over the past few years as a result of the growth in email subscribers. In this growing world, most of the transactions, business, study materials are taking place through emails. But due to the social networks and advertising, some of the emails contain undesirable information known as spam. Reliable anti-spam filters have emerged as a result of the surge in the volume of unsolicited bulk email messages, or spam. Various email spam messages are marketable in nature but might similarly encompass covered links that seem to be for conversant websites but actually lead to phishing web sites or sites that are holding malware. A spammer naturally sends an email to number of email addresses. Spamming is on the rise, thanks to the ease with which it is possible to send a flood of emails at no cost. Spam mails disturb the one calmness. Even though numerous algorithms have been created for the classification of spam emails, none of them consistently produce reliable results. In this study, we are classifying the ham and the spam emails by doing comparative study of the RNN and SVM algorithm. This classification shows 97 and 79% accuracy and 0.97 and 0.79% Kappa coefficient. As a result, this study reveals that the email spam classification of Kaggle datasets has a high level of accuracy.

Abstract

Keywords