EKSPLORASI ARSITEKTUR CONVOLUTIONAL NEURAL NETWORK (CNN) UNTUK VALIDASI INFORMASI DAN DETEKSI RUMOR PADA ARTIKEL BERITA

Abstract

The spread of hoaxes and rumors in digital media has become a serious issue that requires technological solutions for automatic detection. This study aims to explore a text-based news classification system using a Convolutional Neural Network (CNN) architecture to detect information validity and rumors in news articles. The research methodology includes data collection from various online sources, text preprocessing, tokenization, CNN model training, and system testing. The dataset used consists of 1,000 news articles, divided into 500 valid news articles and 500 invalid (hoax) articles. Data were collected from press council–verified news outlets for the valid category and from anonymous sources for the hoax category. Preprocessing steps include text cleaning, tokenization, padding, and label encoding. The CNN model was designed with embedding layers, 1D convolution, global max pooling, dropout, and dense layers for binary classification. The results show that the CNN model achieved an accuracy of 85–90% in classifying valid and invalid news. The model demonstrated good performance in recognizing factual news from credible sources but faced challenges with hoax articles written in formal language styles. The system is capable of making predictions in real time as well as in batch processing through Excel files, making it practical for large-scale implementation.