News classification using light gradient boosted machine algorithm

Abstract

News classification is a complex issue as people are easily convinced of misleading information and lack control over the spread of fake news. However, we ca n break the problem of spreading fake news with artificial intelligence (AI), which has developed rapidly. This study proposes a news classification model using a light gradient boosted machine (LightGBM) algorithm. The model is analyzed using two feature extraction techniques, count vectorizer and Tfidf vectorize r and compared with a deep learning model using long - short term memory (LSTM). The experimental evaluation showed that all LightGBM models outperform LSTM. The best model is the count vectorizer Li ghtGBM, which achieves an accuracy value of 0.9933 and an area under curve (AUC) score of 0.9999.