Implementation of BiLSTM-SVM Algorithm to Detect Fake News on Text-Based Media

Abstract

Online media is one of the places where news can spread quickly and everyone can access it easily and freely. Not only real or valid news is spread on online media, but fake news can also be easily spread on online media, and readers sometimes do not realize that the news they read is fake. As a result, wrong opinions arise that can lead to disputes, as well as divisions between individuals or groups. This study implements the BiLSTM-SVM algorithm to detect fake news that is spread on one of the online media, namely Twitter. The steps taken are tidying up the news text (text preprocessing), converting every word from the news text into numbers in vector form (word embedding), processing the numbers, and then classifying the results of the processing with the BiLSTM-SVM model formed with TensorFlow 2.0 help, and see the performance generated by the BiLSTM-SVM algorithm. The results obtained include an accuracy rate of 86% and an F1 Score value of 87.5% in detecting news from data validation with the same news topic.