A Feature-Based Optimization Approach for Fake News Detection on Social Media Using K-Means Clustering

The social networks and news ecosystem provide valuable social information, however, the rise of deceptive content such as fake news generated by social media users, poses an increasing threat to the propagation and diffusion of fake news over the social network and among users. Low-quality news and misinformation spread on social media had negative impacts on individuals and society. Hence, it is essential to detect fake news to ensure the spread of accurate and truthful information. To address this problem, a new approach using Binary Bat Algorithm (BBA) for fake news detection (FND) on Twitter data is proposed in this paper. Twitter data usually generates massive feature space which might consist of irrelevant features that could jeopardize the subsequent process. The proposed FND approach involves four stages, namely data collection, pre-processing, feature extraction, and fake news detection. The proposed techniques are tested on PHEME dataset, and the experimental results are measured in term average of Precision (PR), Recall (R), F-measure (F), and Accuracy (ACC). The experimental results show that the BBA algorithm has outperformed the Social Spider Optimization (SSO) algorithm. Thus, BBA is a promising solution for solving high-dimensionality feature space in fake news Twitter data.