DefinePK

Abstract

The presence of Short Message Service (SMS) that indicate fraud acts is rising and very disturbing for SMS users which is known as spam SMS. Therefore, it is very important to automatically detect or filter spam SMS. This research developed a system that could classify SMS between SMS spam with not spam (ham) in Bahasa (Indonesian Language). This system conducted with Multinomial Naïve Bayes classification with the feature weighting Term Frequency - Inverse Document Frequency (TF-IDF). Before the classification, data had been preprocessed using tokenization, slang handling, stopword, and stemming. The evaluation is done by using cross validation and were conducted by comparing several test scenarios based on the selected preprocessing technique. From the experiment the best results were obtained 94.44% in accuracy with preprocessing slang handling and stemming. This best result were implemented on the mobile Android with adding rule if the sender of SMS is not in the contact list, then the incoming SMS would be processed to test whether it is spam or ham. From the experiment on Android mobile application accuracy raised until 94.74%.