Table 11.
Comparison of AI-based fake news detection techniques
| Reference | Approach | Method | Model |
|---|---|---|---|
| Del Vicario et al. (2019) | An approach to analyze the sentiment associated with data textual content and add semantic knowledge to it | ML | Linear Regression (LIN), Logistic Regression (LOG), Support Vector Machine (SVM) with Linear Kernel, K-Nearest Neighbors (KNN), Neural Network Models (NN), Decision Trees (DT) |
| Elhadad et al. (2019) | An approach to select hybrid features from the textual content of the news, which they consider as blocks, without segmenting text into parts (title, content, date, source, etc.) | ML | Decision Tree, KNN, Logistic Regression, SVM, Naïve Bayes with n-gram, LSVM, Perceptron |
| Aswani et al. (2017) | A hybrid artificial bee colony approach to identify and segregate buzz in Twitter and analyze user-generated content (UGC) to mine useful information (content buzz/popularity) | ML | KNN with artificial bee colony optimization |
| Hakak et al. (2021) | An ensemble of machine learning approaches for effective feature extraction to classify fake news | ML | Decision Tree, Random Forest and Extra Tree Classifier |
| Singh et al. (2021) | A multimodal approach, combining text and visual analysis of online news stories to automatically detect fake news through predictive analysis to detect features most strongly associated with fake news | ML | Logistic Regression, Linear Discrimination Analysis, Quadratic Discriminant Analysis, K-Nearest Neighbors, Naïve Bayes, Support Vector Machine, Classification and Regression Tree, and Random Forest Analysis |
| Amri et al. (2022) | An explainable multimodal content-based fake news detection system | ML | Vision-and-Language BERT (VilBERT), Local Interpretable Model-Agnostic Explanations (LIME), Latent Dirichlet Allocation (LDA) topic modeling |
| Wang et al. (2019b) | A hybrid deep neural network model to learn the useful features from contextual information and to capture the dependencies between sequences of contextual information | DL | Recurrent and Convolutional Neural Networks (RNN and CNN) |
| Wang (2017) | A hybrid convolutional neural network approach for automatic fake news detection | DL | Recurrent and Convolutional Neural Networks (RNN and CNN) |
| Liu and Wu (2018) | An early detection approach of fake news to classify the propagation path to mine the global and local changes of user characteristics in the diffusion path | DL | Recurrent and Convolutional Neural Networks (RNN and CNN) |
| Mishra (2020) | Unsupervised network representation learning methods to learn user (node) embeddings from both the follower network and the retweet network and to encode the propagation path sequence | DL | RNN: (long short-term memory unit (LSTM)) |
| Qian et al. (2018) | A Two-Level Convolutional Neural Network with User Response Generator (TCNN-URG) where TCNN captures semantic information from the article text by representing it at the sentence and word level. The URG learns a generative model of user responses to article text from historical user responses that it can use to generate responses to new articles to assist fake news detection | DL | Convolutional Neural Network (CNN) |
| Zhang et al. (2020) | Based on a set of explicit features extracted from the textual information, a deep diffusive network model is built to infer the credibility of news articles, creators and subjects simultaneously | DL | Deep Diffusive Network Model Learning |
| Goldani et al. (2021) | A capsule networks (CapsNet) approach for fake news detection using two architectures for different lengths of news statements and claims that capsule neural networks have been successful in computer vision and are receiving attention for use in Natural Language Processing (NLP) | DL | Capsule Networks (CapsNet) |
| Wang et al. (2019b) | An automated approach to distinguish different cases of fake news (i.e., hoaxes, irony and propaganda) while assessing and classifying news articles and claims including linguistic cues as well as user credibility and news dissemination in social media | DL, ML | Convolutional Neural Network (CNN), long Short-Term Memory (LSTM), logistic regression |
| Abdullah-All-Tanvir et al. (2019) | A model to recognize forged news messages from twitter posts, by figuring out how to anticipate precision appraisals, in view of computerizing forged news identification in Twitter dataset. A combination of traditional machine learning, as well as deep learning classification models, is tested to enhance the accuracy of prediction | DL, ML | Naïve Bayes, Logistic Regression, Support Vector Machine, Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM) |
| Kaliyar et al. (2020) | An approach named (FNDNet) based on the combination between unsupervised learning algorithm GloVe and deep convolutional neural network for fake news detection | DL, ML | Deep Convolutional Neural Network (CNN), Global Vectors (GloVe) |
| Zhang et al. (2019a) | A hybrid approach to encode auxiliary information coming from people’s replies alone in temporal order. Such auxiliary information is then used to update a priori belief generating a posteriori belief | DL, ML | Deep Learning Model, Long Short-Term Memory Neural Network (LSTM) |
| Deepak and Chitturi (2020) | A system that consists of live data mining in addition to the deep learning model | DL, ML | Feedforward Neural Network (FNN) and LSTM Word Vector Model |
| Shu et al. (2018a) | A multidimensional fake news data repository “FakeNewsNet” and conduct an exploratory analysis of the datasets to evaluate them | DL, ML | Convolutional Neural Network (CNN), Support Vector Machines (SVMs), Logistic Regression (LR), Naïve Bayes (NB) |
| Vereshchaka et al. (2020) | A sociocultural textual analysis, computational linguistics analysis, and textual classification using NLP, as well as deep learning models to distinguish fake from real news to mitigate the problem of disinformation | DL, NLP | Short-Term Memory (LSTM), Recurrent Neural Network (RNN) and Gated Recurrent Unit (GRU) |
| Kapusta et al. (2019) | A sentiment and frequency analysis using both machine learning and NLP in what is called text mining to processing news content sentiment analysis and frequency analysis to compare basic text characteristics of fake and real news articles | ML, NLP | The Natural Language Toolkit (NLTK), TF-IDF |
| Ozbay and Alatas (2020) | A hybrid approach based on text analysis and supervised artificial intelligence for fake news detection | ML, NLP | Supervised algorithms: BayesNet, JRip, OneR, Decision Stump, ZeroR, Stochastic Gradient Descent (SGD), CV Parameter Selection (CVPS), Randomizable Filtered Classifier (RFC), Logistic Model Tree (LMT). NLP: TF weighting |
| Ahmed et al. (2020) | A machine learning and NLP text-based processing to identify fake news. Various features of the text are extracted through text processing and after that those features are incorporated into classification | ML, NLP | Machine learning classifiers (i.e., Passive-aggressive, Naïve Bayes and Support Vector Machine) |
| Abdullah-All-Tanvir et al. (2020) | A hybrid neural network approach to identify authentic news on popular Twitter threads which would outperform the traditional neural network architecture’s performance. Three traditional supervised algorithms and two Deep Neural are combined to train the defined model. Some NLP concepts were also used to implement some of the traditional supervised machine learning algorithms over their dataset | ML, DL, NLP | Traditional supervised algorithm (i.e., Logistic Regression, Bayesian Classifier and Support Vector Machine). Deep Neural Networks (i.e., Recurrent Neural Network, Long Short-Term Memory LSTM). NLP concepts such as Count vectorizer and TF-IDF Vectorizer |
| Kaur et al. (2020) | A hybrid method to identify news articles as fake or real through finding out which classification model identifies false features accurately | ML, DL, NLP | Neural Networks (NN) and Ensemble Models. Supervised Machine Learning Classifiers such as Naïve Bayes (NB), Decision Tree (DT), Support Vector Machine (SVM), Linear Models. Term Frequency-Inverse Document Frequency (TF-IDF), Count-Vectorizer (CV), Hashing-Vectorizer (HV) |
| Kaliyar (2018) | A fake news detection approach to classify the news article or other documents into certain or not. Natural language processing, machine learning and deep learning techniques are used to implement the defined models and to predict the accuracy of different models and classifiers | ML, DL, NLP | Machine Learning Models: Naïve Bayes, K-nearest Neighbors, Decision Tree, Random Forest. Deep Learning Networks: Shallow Convolutional Neural Networks (CNN), Very Deep Convolutional Neural Network (VDCNN), Long Short-Term Memory Network (LSTM), Gated Recurrent Unit Network (GRU). A combination of Convolutional Neural Network with Long Short-Term Memory (CNN-LSTM) and Convolutional Neural Network with Gated Recurrent Unit (CNN-LSTM) |
| Mahabub (2020) | An intelligent detection system to manage the classification of news as being either real or fake | ML, DL, NLP | Machine Learning: Naïve Bayes, KNN, SVM, Random Forest, Artificial Neural Network, Logistic Regression, Gradient Boosting, AdaBoost |
| Bahad et al. (2019) | A method based on Bi-directional LSTM-recurrent neural network to analyze the relationship between the news article headline and article body | ML, DL, NLP | Unsupervised Learning algorithm: Global Vectors (GloVe). Bi-directional LSTM-recurrent Neural Network |