查看文章

The increasing number of cell phone users guides to rising of SMS spam messages. Facing cell phone spams is hard because of several reasons such as spam filtering systems adapted in the cell phone are limited, and lower rate of these types of messages actually leads users and providers to pay no attention to spam filtering problem. Spam filtering applied in Short Messaging Services should use one of the text classification methods to classify SMS either as a spam or ham message. This paper investigates 12 different SMS classifiers. Among of these classifiers, Discriminative Multinomial Naïve Bayes, Stochastic Gradient Descent, and Support Vector Machine show the highest accuracy. As a result of including two equally ham and spam SMS, we conclude that Discriminative Multinomial Naïve Bayes gains the highest accuracy 96.46%, then SGD with 96.13%, whereas SVM has 95.87%., while using the whole data set with the same classifiers, the results are promising and SVM shows accuracy of 98.6%.