Abstract

SMS Spam is an unsolicited or unwanted text message by a user that is sent to a mobile device.At this time, increasingly criminal acts can annoy recipients by spreading unsolicited or unwanted spam SMS, including promotions, fraud, pornographic messages, and others.Therefore, the classification of SMS needs to be developed to assist in categorizing SMS.In existing research, to try to overcome these problems, the term frequency-inverse document frequency (TF-IDF) feature is applied.However, this method has a disadvantage, namely eliminating category information on each document, so in this study, a comparison will be made with the Supervised Term Weighting feature method, which is one of the terms frequency relevance frequency (TF-RF) using the Support Vector Machine, K-nearest Neighbor, and Multinomial Naïve Bayes.The total data used is 500 SMS with a comparison of 325 non-spam SMS and 175 spam SMS.After the experiment is conducted, SVM Kernel Sigmoid has the highest average accuracy value where the difference in average accuracy with Kernel RBF is 2.26%, Linear Kernel is 0.09%, k-Nearest Neighbor is 27.56%, and Multinomial Naïve Bayes is 4.37%.

Concepts :

Access to Document

Citations by Year