Spam Classification with Support Vector Machines Using Van der Waerden Rank Score Attention

Area of Science:

• Computer Science
• Machine Learning
• Network Security

Background:

• Spam poses a significant threat to network security and communication efficiency.
• Conventional spam detection methods like traditional machine learning and deep learning have limitations in handling high-dimensional data or computational resources.

Purpose of the Study:

• To introduce an efficient and interpretable spam classification method.
• To address the limitations of existing spam detection techniques.

Main Methods:

• A novel Van der Waerden rank score feature attention-enhanced Support Vector Machine (VWR-Attn-SVM) was developed.
• Van der Waerden rank transformation was used for text feature normalization, enhancing outlier robustness and preserving ordinal relationships.
• An enhanced attention mechanism with non-linear processing and regularization was employed for optimized feature selection.

Main Results:

• VWR-Attn-SVM demonstrated superior performance over traditional classifiers on the UCI Spambase and Indonesian Spam datasets.
• The method achieved higher accuracy, precision, recall, F1-score, and AUC.
• The approach combines high performance with reduced computational cost.

Conclusions:

• VWR-Attn-SVM provides an efficient and interpretable solution for spam classification.
• The method shows potential for application in other text-based platforms like messaging and social media.
• This technique offers a promising advancement in combating spam effectively.