Development of Short Message Service Spam Filtering Using Naive Bayes Algorithm

Abstract

Short Message Service (SMS) is a text message service available in smart-phones as well as basic phones. The daily increase in the number of mobile device users drastically increases SMS traffic which in turns increases the number of spam messages. Spam is a very serious universal problem that causes a lot of damages. Several studies have been presented, including implementations of spam filters to prevent spam from reaching their destination. Mobile spam is a growing problem that keeps increasing day by day. The complexity of the messages that spammers impose has made it harder to classify spam. The difficulty in sifting through SMS spam is that messages are typically short, with phrases made up of acronyms and abbreviations, which makes them more ambiguous. In this paper, a supervised machine learning algorithm called Naïve Bayes Algorithm is used as an effective technique in SMS spam filtering. The dataset used for training and testing models was downloaded from UCI machine learning repository that contains 5574 English raw text messages; where 4827 are ham messages and 747 are spam messages. The results obtained show accuracy score of 98.7%, precision 96.1%, recall 94.5% and F1-score 95.3%. The experimental results have shown that, Naïve Bayes Machine learning model performs better than other supervised machine learning models during the training and testing of SMS spam detection and filtering.