A STUDY COMPARING MACHINE LEARNING AND DEEP LEARNING METHODS FOR THE DISCOVERY OF FAKE NEWS IN HEALTH SECTOR USING DIFFERENT DEEP AND SHALLOW METHODS

Abstract

ABSTRACT The problem of fake news, which existed even before Internet prevalence, has been made worse by the internet's growth and adoption. If the news is concerning your health, this becomes more urgent. This study suggests Content Based Models (CBM) and Feature Based Models (FBM) as solutions to this problem. The supplied input is what distinguishes the two models. The FBM also takes two readability features as input in addition to the content, whereas the CBM simply accepts news content as an input. Under each category, the effectiveness of two hybrid deep learning approaches, namely CNN-LSTM and CNN- BiLSTM, is compared with five classic machine learning techniques: Decision Tree, Random Forest, Support Vector Machine, AdaBoost-Decision Tree, and AdaBoost-Random Forest. The study used the Fake News Healthcare dataset, which included 9581 stories. This extremely unbalanced dataset is balanced using a simple data augmentation technique. The experimental findings show that Feature Based Models outperform Content Based Models in terms of performance. AdaBoost- Random Forest had an F1 Score of 98.9%, while the Hybrid CNN-LSTM model had an F1 Score of 97.09% among the proposed FBM. The best-performing model for classifying fake news is Adaboost-Random Forest under FBM.