A Deep Learning Approach to Classify Spam URLs Data using MLP
Jabeen Sultana, Tamanna Siddiqui · 2025
Spam emails pose a growing threat to networks and their users despite the availability of various detection techniques. With the increase in browsing usage, there is a greater risk that URLs will become prone to spam. This research introduces a deep learning method that combines a traditional approach using enhanced activation functions to attain efficient classification results. This suggested approach classifies spam URLs into 2-class two-classifications. This research work improves the accuracy of classification while reducing misclassification rates. The dataset is preprocessed well, trained, and tested on four different classifier models, achieving a maximum classification accuracy of 96.99% and a minimum of 80.26% for spam email classification. The AUC_ROC scores ranged from 0.92 to 0.43, highlighting the model's robustness. Additionally, the model efficiently processes large datasets while maintaining high classification accuracy, as demonstrated through key metrics such as accuracy, precision, recall, F-Score, and ROC scores.