KOMPARASI METODE SUPPORT VECTOR MACHINE (SVM) DAN RECURRENT NEURAL NETWORK (RNN) UNTUK KLASIFIKASI SPAM SMS

Short Message Service or commonly called short message SMS is part of the communication message service on cellular phones. However, SMS are now rarely used because their functions have been replaced by chat applications such as Whatsapp, Line, Telegram, etc. However, the SMS was not removed for one reason, namely the SMS used by various applications to verify the code as a sign that the phone number used was active. Increasing use of SMS is widely misused by certain parties to commit crimes such as fraud, spreading hoaxes, and sending phishing links that are vulnerable to cyber via SMS. To classify SMS spam, a sentiment analysis technique is needed to find out whether the SMS sentence is spam or ham. This study aims to compare the predictions of each spam SMS uses the Support Vector Machine (SVM) and Recurrent Neural Network (RNN) methods. The data used are 5,559 with the category of 747 spam texts and 4,812 ham texts based on data taken from Kaggle.com. Data SMS spam modeled using the Support Vector Machine (SVM) method with weighting the method using Term Frequency – Inverse Document Frequency (TF-IDF) and by doing several tests with split data testing as much as 0.2 (20%), 0.3 (30%), and 0.4 (40%) obtained the best accuracy of 0.99 (99%) using data testing 0.2 (20%). In the Recurrent Neural Network (RNN) method, several tests were carried out with split data testing as much as 20%, 30%, and 40% and the best accuracy results were obtained at epoch 15 of 0.9988 (99.88%) using data testing 0.3 (30%)