Abstract:
The rapid expansion of misinformation in daily life has disrupted different news sources, such as social media, online news, radio and television stations, and newspapers, making it difficult to choose reliable news outlets. The potential to spread fake news (FN) to many organizations and platforms jeopardizes news credibility and causes users to abandon them. However, detecting FN entails predicting the probability that a particular news article is deceptive or not. However, most contemporary methods do not consider Arabic news and how Arabic FN (AFN) has been detected in the past decade. Therefore, research on AFN detection is beginning to receive more attention. This paper presents an Arabic FN detection (AFND) system based on hybrid deep learning (DL) model. This model includes both conventional neural network and long short-term memory (CNN-LSTM) modalities. The input dataset was prepared via discretization and normalization. Then, word vectors were included with the corrected words at a given word length as pretrained vectors on Arabic news. Due to outstanding performance, the JSO optimization algorithm was combined with the framework to automatically define the best structure for the proposed CNN-LSTM. A comparison was made between the proposed CNN-LSTM and other recent models to prove the performance of the proposed CNN-LSTM. The results indicate that the proposed CNN-LSTM offers the best performance, with an accuracy of 81.6%. The experimental results provided comprehensive improvements in the subject matter of AFND and demonstrated the potential of the proposed methodology.
Page(s):
5072-5086
DOI:
DOI not available
Published:
Journal: Journal of Theoretical and Applied Information Technology, Volume: 100, Issue: 14, Year: 2022
Keywords:
optimization
,
Deep Learning DL
,
ensemble learning
,
Conventional Neural Network CNN
,
Arabic Fake News Detection AFND
,
Long ShortTerm Memory LSTM