Fake News Prediction Using Hybrid Model–Systematic Literature Review

The proliferation of fake news is recognized as a deliberate strategy to manipulate and misinform audiences, thereby undermining the consumption of authentic information. The repercussions of this phenomenon span from mere annoyance to the significant potential for distorting societal and even national perspectives. The contemporary technological landscape has expedited the dissemination of information, underscoring the urgency of discerning evolving techniques in fake news detection. In the context of prevalent social media platforms, a sole reliance on content-based methodologies proves inadequate. This study employs a systematic literature review following the PRISMA protocol to illuminate the contemporary landscape of fake news detection methodologies. The investigation reveals a spectrum of strategies categorized under the rubric of hybrid models, wherein multiple features or models are amalgamated. The discerned hybrid models exhibit a diversity of methodologies, coalescing into two overarching paradigms: the fusion of diverse features and the integration of multiple models. The former primarily encompasses composite feature-based ensembles, often amalgamating content-based features with complementary attributes. The latter paradigm predominantly entails the synthesis of various deep learning approaches, culminating in enhanced performance metrics for fake news detection. The synthesized findings not only provide a comprehensive overview of the prevalent hybrid approaches but also establish a benchmark for forthcoming researchers embarking on predictive endeavors involving hybrid methodologies.