Enhancing Spam Email Detection with a Soft Voting Ensemble of Optimized Machine Learning

Spam email detection is crucial for cybersecurity, as it protects user privacy and reduces security risks. The persistent presence of spammers necessitates continuous improvements in spam filtering measures. To address this challenge, this study employs Grid Search Optimizer to fine-tune the parameters of four distinct classifiers: Support Vector Machine (SVM), Random Forest, Naive Bayes, and XGBoost. Soft Voting serves as the final classifier, enhancing overall accuracy by integrating results from optimized classifiers to detect and classify spam emails across the Spam Mails Dataset and Enron1 Dataset. The experimental study compared the proposed model with relevant works and found that hyperparameter tuning and Soft Voting significantly enhanced its performance compared to existing approaches. The proposed ensemble model outperformed individual classifiers, achieving accuracies of 99.32% and 99.12%, respectively. Also, the ensemble model demonstrated an AUC of 1.00 on both datasets, indicating its effectiveness in distinguishing between spam and ham emails. This approach surpasses prior studies in accuracy, generalization, and robustness by innovatively combining Grid Search and Soft Voting. The constructed model exhibits high effectiveness and efficiency in the detection of spam emails.