A Generalized Two-Level Ensemble Method for Spam Mail Detection

Abstract

Email is the most cost-effective way to communicate with people across the world. It offers a simple and convenient way to send and receive messages. However, it is susceptible to various types of threats. The most significant risk to emailers is spam, which refers to unsolicited emails that are often sent in large numbers to multiple recipients. Malicious spam includes links to phishing websites. These links not only pose a threat to our system, but also make our personal information accessible to hackers. Spammers often create fake profiles and email accounts making it easier for them to deceive unsuspecting victims. In this paper, the content of the email is used to detect spam which provides more information than a URL. It has been found that several techniques have been proposed to efficiently identify spam emails but current email spam detection methods have not yet achieved high accuracy. It is necessary to improve their performance in spam detection, a generalized two-level ensemble technique is explored irrespective of the type of spam datasets. The proposed two-level ensemble algorithm compares with various machine learning techniques such as SVM, Regression, and kNN, along with their variants, Finally, a K-fold and voting algorithms are applied to make the final prediction.