Optimizing Arabic Spam Filtering Through Unsupervised and Ensemble Learning Approaches

EasyChair Preprint 11608

Abstract

This paper presents a new Ensemble Learning approach for filtering Arabic spam. The proposed approach utilizes four unsupervised Machine Learning algorithms, including One Class Support Vector Machine (OCSVM), the Histogram-Based Outlier Score (HBOS), Local Outlier Factor (LOF) and Isolation Forest (IF), to construct a robust spam filter. The performance of our proposed approach is evaluated on a textual Arabic dataset.
The experimental results show that our model achieves more than 84% of accuracy outperforming other Machine Learning algorithms. The use of Ensemble Learning and multiple unsupervised algorithms in our approach proves to be a promising solution for effective Arabic spam filtering.

Keyphrases: Arabic Spam, ensemble learning, learning, machine learning, unsupervised

@booklet{EasyChair:11608,
  author    = {Marouane Kihal and Lamia Hamza},
  title     = {Optimizing Arabic Spam Filtering Through Unsupervised and Ensemble Learning Approaches},
  howpublished = {EasyChair Preprint 11608},
  year      = {EasyChair, 2023}}