Implementasi Metode LSTM Untuk Mengklasifikasi Berita Palsu Pada PolitiFact

Abstract

Fake news is an increasingly worrying issue in the digital age, as its spread has the potential to impact public perception and the democratic process. The information classified in this study consists of political claims assessed by Politifact, along with Politifact's evaluation of their accuracy. The purpose of this research is to create a fake news classification system concentrated on data from Politifact using Long Short-Term Memory (LSTM). The methods to be used include data collection from the PolitiFact.com website that provides fake news and fact news, followed by data labeling, Exploratory Data Analysis (EDA), data preprocessing, use of wordcloud visualization, data separation, LSTM model formation, evaluation, testing with new data to be classified. The data used amounted to 1500 news with the number of fact news as many as 34 and fake news as many as 1466. The results showed that the LSTM model was very good at classifying fake news in Politifact by being able to produce a very high accuracy rate of 97%. The ability of this model can be one of the right choices for classifying fake news and can be an invaluable tool in combating the spread of false information and supporting reliable media in disseminating accurate news to the public.