Active Learning Approach for Spam Filtering

Abstract

It is general to relate an optimization problem with bound constraints and a linear equality constraint with training a support vector machine (SVM). In this paper, the Support Vector Machine (SVM) model trained by the Active Learning algorithm is proposed for spam filtering. This experiment proceeds the data preprocessing, training of the SVM model, application of Active Learning (AL), evaluation of performances, and comparisons of results between random sampling and AL. The graphics clearly exhibit the result that the Active Learning method could use only a small size of the dataset to achieve efficient and accurate spam filtering work. This experiment assists in enhancing the efficiency of working and processing emails.