Spam Detection Using Clustering-Based SVM

Abstract

Spam detection task is of much more importance than earlier due to the increase in the use of messaging and mailing services. Efficient classification in such a variety of messages is a comparatively onerous task. There are a variety of machine learning algorithms used for spam detection, one of which is Support Vector Machine, also known as SVM. SVM is widely used to classify text-based documents. Though SVM is a widely used technique in document classification, its performance in the spam classification is not the best due to the uneven density of the training data. In order to improve the efficiency of SVM, I introduce a clustering-based SVM method. The training data is pre-processed using clustering algorithms and then the SVM classifier is implemented on the processed dataset. This method would increase the performance by overcoming the problem of uneven distribution of training data. The experimental results show that the performance is improved compared to that of SVM.