Feature Fusion Strategies for Robust Fake News Detection: A Hybrid Deep Learning Approach with CBF and ABF Models

Read full text Summarise with VIRAT
Authors: Poonam S. Jadhav, Sarika Nitin Zaware
Cited by: 0
References: 18

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

The dissemination of fake news is a critical threat to an individual and community, and its identification is a significant subject of study.Although several automated detection systems have been created, most of them are not normally in a position to correctly categorize text-based fake news.The complexity of this is hard to model using the traditional approaches that rely on linguistic, semantic, syntactic or statistical aspects.In order to address these acute problems, this study suggests a new feature fusion algorithm to improve the functionality and stability of the fake news detection.In our model, we combine linguistic features, syntactic features, embedded in context features and are further supported by Timex feature to obtain pertinent time context.Leaky feature removal is done before the training to guarantee a generalization of the model as well as to avoid biased learning, evaluation of the model is done through a k-fold cross-validation approach.The polished feature set is processed by deep learning architectures, concatenationbased fusion (CBF) and attention-based fusion (ABF) is used to improve its feature integration and classification and the proposed model is systematically compared to traditional machine learning algorithms.The proposed CBF approach is proved to be effective, based on results of experiments with the LIAR-PLUS (multi-class) dataset, where the accuracy of the approach is 91.2% and 94.8% when the ABF strategy is applied.The results also have a relatively low rate of false-positive compared to the existing strategies.These features have been evaluated as significant and it should be noted that the fusion strategy and combo of features yields better results in terms of fake news detection.On the whole, the present study does not only summarize the existing situation in fake news detection but it also gives a strong foundation to further developments in detecting textual misinformation.

Fields of Study & Subjects

Misinformation and Its Impacts Spam and Phishing Detection Hate Speech and Cyberbullying Detection

Publication details

Journal / source
International journal of intelligent engineering and systems
Volume
19
Issue
4
Pages
620–638
Publication date
23 March 2026
Language
EN