Implementation and evaluation of AI Models For fake news classification on Social Media

Abstract

The fast development of different social media has simplified the dissemination of fake news to the extreme, and endangered the credibility of information, popular opinion, and social order seriously. The present study presents the introduction of implementation and assessment of various artificial intelligence (AI)-based systems to classify fake news in social media posts. The research looks at both classic “machine learning algorithms and advanced deep learning architectures to” make readings of the textual characteristics of news post and user generated content. Data preprocessing techniques such as tokenizing, stop- word elimination, and lemmatization as well as vectorizing the data were employed to “improve the quality of the data” as well as model performance. The models were trained and tested on some benchmark fake news datasets. They were subsequently tested in terms of some standard metrics of performance like accuracy, precision, recall and F1-score and also confusion matrix analysis. The results of the tests show that deep learning models, particularly those of transformer and recurrent neural network architecture are more efficient than the old machine learning techniques to comprehend the contextual and semantic facet of the provided text. However, one can regard traditional models as a close competitor of deep learning ones since they are less complicated to compute, thus, more suitable to be used in real-time.