Comparative Analysis of Machine Learning Classifiers for E-mail Spam Filtering with RFE-SVM Feature Selection and SA-Optimised XGBoost

• Abubakar HASSAN

  Author

• Emmanuel G. DADA

  Author

This study presents a comparative evaluation of several machine learning classifiers for email spam filtering using a Kaggle dataset of 5,172 email samples. The investigation includes a proposed hybrid approach combining RFE-SVM feature selection with SA-optimized XGBoost, evaluated alongside SVM, MLP, Gaussian Naïve Bayes, and Logistic Regression. Results showed that MLP achieved the highest accuracy at 99.2%, followed by SVM at 99.0%, while the proposed SA-XGBoost achieved 98.3% with notably lower memory consumption of 0.20 MB. The findings suggest that for this dataset, simpler classifiers may outperform more complex ensemble methods, and that the proposed approach offers a favourable accuracy-memory trade-off for resource-constrained environments.

• 76_Hassan and Dada (2026)

Copyright (c) 2026 Abubakar HASSAN, Emmanuel G. DADA (Author)

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.