Using One-Class SVM with Spam Classification

Authors

  • Inas Ali Department of Computer, College of Science, University of Baghdad, Baghdad, Iraq
  • Sumaya Saad Department of Computer, College of Science, University of Baghdad, Baghdad, Iraq
  • Safa Ahmed Department of Computer, College of Science, University of Baghdad, Baghdad, Iraq

Keywords:

gain ratio, spam, SVM

Abstract

Support Vector Machine (SVM) is supervised machine learning technique which has become a popular technique for e-mail classifiers because its performance improves the accuracy of classification. The proposed method combines gain ratio (GR) which is feature selection method with one-class training SVM to increase the efficiency of the detection process and decrease the cost. The results show high accuracy up to 100% and less error rate with less number of feature to 5 features.

Connection failed: User ijs already has more than 'max_user_connections' active connections