International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 06 | Jun 2025
p-ISSN: 2395-0072
www.irjet.net
Fake News v/s Satire: Detection and Classification Chinmay Deodhar 1, Soham Kulkarni 2 1 Department of Information Dwarkadas J, Sanghvi College of Engineering, Mumbai, India
2 Department of Information Dwarkadas J, Sanghvi College of Engineering, Mumbai, India --------------------------------------------------------------------***----------------------------------------------------------------------(BERT). In this paper, we compare the performance of Abstract: In recent years, the proliferation of fake news
several machine learning models in detecting and classifying fake news and non-fake news articles. The models are trained and evaluated on a dataset of news articles manually labelled as either fake or non-fake. We evaluate the performance of the following models: SVM, Multinomial Naive Bayes, Random Forest, XGBoost, CNN, RNN, LSTM, and BERT. The results of this study can provide insights into the most effective machine learning models for detecting and combating fake news and misinformation. In particular, our findings suggest that ensemble methods and deep learning models may be more effective in detecting fake news and non-fake news articles than traditional machine learning models.
and misinformation has become a significant problem, creating a need for accurate and efficient detection and classification methods. This research paper compares the performance of several machine learning models, including Support Vector Machines (SVM), Multinomial Naive Bayes, Random Forest, XGBoost, Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), and Bidirectional Encoder Representations from Transformers (BERT) in detecting and classifying fake news and non-fake news articles. The models were trained and evaluated on a dataset of news articles labelled as either fake or non-fake. Our results show Random Forest and CNN outperform the other models, achieving an accuracy of 94% and 94.25%, respectively. SVM and RNN also achieved reasonably high accuracy, with accuracies of 89% and 93.64%, respectively. The accuracies of Multinomial Naive Bayes, LSTM, and XGBoost were 86%, 80%, and 61%, respectively. The lower accuracy of BERT (61%) is attributed to training the model on a smaller dataset due to limited computing resources. Our findings suggest that Random Forest and CNN are promising models for detecting and classifying fake news and non-fake news articles. They can be used to develop effective solutions for combating fake news and misinformation.
Literature Review A number of studies have investigated the effectiveness of different machine learning techniques in detecting fake news. For example, Golbeck et al. [1] created a dataset of fake news and satire articles and tested the accuracy of different classifiers, including SVM, Naive Bayes, and Random Forest. They found that Random Forest outperformed the other classifiers with an accuracy of 94%. Similarly, Khan et al. (2021) [3] compared the performance of several machine learning models, including SVM, Naive Bayes, and Deep Learning, and found that Deep Learning achieved the highest accuracy of 92%.
Keywords: fake news, satire, misinformation, natural language processing, CNN, BERT, transformers
Another area of research has focused on the use of linguistic and semantic cues to differentiate between fake news and satire. Levi et al. [2] explored this approach by analyzing the use of language patterns in fake news and satire articles. They found that while both types of articles often contained exaggeration and mockery, fake news articles tended to use more hyperbolic language and appeal to emotion, while satire articles relied more on irony and sarcasm.
Introduction Fake news and misinformation have become increasingly prevalent in the digital age. Social media platforms and the internet make it easy for false information to spread quickly, leading to serious consequences such as influencing public opinion and affecting elections. The distinction between fake news and non-fake news articles, such as satire or opinion pieces, can be blurred, making it challenging to detect and classify them accurately. In recent years, there have been significant efforts to develop effective solutions for detecting and combating fake news. Machine learning models have been proposed as a promising approach for detecting and classifying fake news articles. These models use various features of the articles, such as language patterns and metadata, to differentiate between real and fake news. A wide range of models has been proposed in the literature, including Support Vector Machines (SVM), Random Forest, XGBoost, Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), and Bidirectional Encoder Representations from Transformers
© 2025, IRJET
|
Impact Factor value: 8.315
In addition to analyzing the content of articles, researchers have also investigated the role of social media in spreading fake news. Asr and Taboada [4] noted that social media platforms are often the primary means by which fake news is disseminated and highlighted the importance of using big data analytics to detect patterns of misinformation. Liu et al. [7] specifically looked at the detection of satirical news on social media and found that using features such as the use of emoticons and hashtags could improve the accuracy of fake news detection. Other studies have explored the use of more sophisticated techniques, such as deep learning algorithms, in detecting
|
ISO 9001:2008 Certified Journal
|
Page 924