SPAM DETECTION FOR EMAIL FILTERS USING ARTIFICIAL INTELLIGENCE TECHNIQUES

Abstract

Spam emails represent a significant threat to digital communications, compromising user privacy and security. Technological advancements have made traditional filtering methods, such as blacklists, conventional machine learning classifiers, and rule-based methods, incapable of adapting to the sophisticated techniques of spammers. This research focuses on providing a systematic analysis of techniques and models used in spam detection. It also examines how these models, such as artificial neural networks (ANNs), extract text features using TF-IDF and classify them to capture complex and nonlinear patterns in email data. In this research, it was found that some of the models proposed by the researchers for spam detection outperform traditional classifiers. The study demonstrated that the hybrid model proposed by the researchers, combining natural language processing and artificial neural networks, exhibits superior performance. When TF-IDF-based feature extraction is combined with an artificial neural network, this approach achieves higher accuracy and more robust semantic representation, enabling it to detect complex linguistic patterns common in sophisticated spam messages. Although the hybrid model requires higher computational costs and greater sensitivity to linguistic variation and class imbalances, its superiority over all other techniques justifies this trade-off. In contrast, models such as artificial neural networks alone, simple Bayesian algorithms, and support vector machines demonstrate acceptable performance but remain limited by either weak semantic understanding or overly robust modelling assumptions.