Explorative Analysis on Email Spam Filtering Sampling a Deep learning and non-deep learning algorithm

ABSTRACT

Email spam includes unwanted emails with tendency of causing harm to the receiver. Addressing this gave rise to spam detection and filtering. While several deep learning and non-deep learning algorithms have been applied for email spam filtering with different data for each of the research, this paper is geared towards combining most of these datasets that produced efficient results from previous researchers and classifying them using the best algorithm for deep and non-deep learning. After carefully reviewing previous works done by other researchers, we adopted Artificial Neural Network (ANN) and Support Vector Machine (SVM) for this research. We also combined the following data for this work; lingspam, completeSpamAssassin, spam_ham, Phishing_Email, and enronSpam. The combination of these sets of data makes the work a unique one. This work, analyzed the deep learning (ANN) and non-deep learning techniques (SVM) for email spam filtering to propose the algorithm that presents the best results for email spam filtering when these datasets are combined. Non-deep learning provided higher accuracy than the deep learning algorithm with Support Vector Machine having precision (0.99), recall (0.98), f1-score (0.98) for ham, and accuracy of 0.98.

Keywords:
Email spam filtering, deep learning, non-deep learning, Artificial Neural Network, Support Vector Machine.