A COMPARATIVE REVIEW OF MACHINE LEARNING TECHNIQUES FOR EMAIL SPAM DETECTION: FROM CLASSICAL MODELS TO MULTIMODEL DEEP LEARNING
Keywords:
Spam Detection, Naive Bayes, Logistic Regression, Random Forest, Machine Learning, Email Forensics, Deep Learning, Comparative ReviewAbstract
Email spam has been a constant menace in the internet space, as it has grown to be more than texts scam- ming but composed of multi-content attacks. Within the last twenty years, spam detection has undergone different eras of development methodologically; starting with rule-based filters, and then evolving to classical machine learning, and most recently to deep learning. In this paper, an analysis of the main techniques has been done and three underlying algorithms, namely Naive Bayes, Logistic Regression, and Random Forest, and their performance, constraints, and their usefulness in real- world settings have been provided. We evaluate their theoretical foundations, feature engineering specifications and their relative accuracy on benchmarks such as SpamAssassin and Enron. In addition, we discuss the way in which the current methods, such as ensemble methods and multi-model deep learning, overcome the limitations of classical models. In this review, we have conducted the synthesis of the 40 recent studies (2013-2025) and concluded that the classical methods are still useful due to their simplicity and interpretability, whereas hybrid methods and transformer-based systems currently represent the state of the art in forensic-grade spam detection.
