Optimised content-social based features for fake news detection in social media using text clustering approach
The task of detecting fake news is highly important to mitigate the misleading information spreading. The accuracy of fake news detection models relies mainly on the quality of the extracted features and the method used in detection. The detection accuracy of these models is still low due to the lac...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | en en en |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://etd.uum.edu.my/12040/1/depositpermission.pdf https://etd.uum.edu.my/12040/2/s902322_01.pdf https://etd.uum.edu.my/12040/3/s902322_02.pdf https://etd.uum.edu.my/12040/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The task of detecting fake news is highly important to mitigate the misleading information spreading. The accuracy of fake news detection models relies mainly on the quality of the extracted features and the method used in detection. The detection accuracy of these models is still low due to the lack of utilizing different features and the limitations in the used methods. To address this problem, this research presents a fake news detection method that combines different features and developing methods for the most crucial phases in the detection process, including feature extraction, feature selection and fake news detection phases. This thesis tackles the extraction process of content-based and social-based features from the news and combines them for better analysis and detection of fake news. In addition, this thesis tackles the feature selection problem by designing a novel wrapper feature selection method based on the Hybrid Flower Pollination Algorithm (HFPA). In particular, this study hybridized the Particle Swarm Optimization algorithm (PSO) with FPA to overcome the limitation of the exploration problem of FPA. In general, the process of fake news detection was conducted in two different phases, the topic detection phase using a graph-based unsupervised clustering method based on HFPA and Markov Clustering Algorithm (MCL) called (HFPA-MCL) and the fake news detection phase using an unsupervised clustering method based on K-means algorithm. The developed fake news detection model is tested on the PHEME dataset by conducting several experiments against different baseline methods. The empirical results of several improvements prove that applying HFPA-MCL on the combination of the optimized content-social features outperforms other methods in terms of feature reduction and detection performance and time. Ultimately, this research concludes that the developed fake news detection model based on the designed content-social-based features can effectively detect fake news in real-world news |
|---|