Table 3 Summary of models and their descriptions used in fake news Detection.
Approach | Strengths | Limitations | Use Cases | Complexity | Interpretability | Scalability | Performance |
|---|---|---|---|---|---|---|---|
Naive Bayes, SVM | Fast, interpretable, simple to implement | Limited semantics, poor generalization | Basic classification, small-scale datasets | Low | High | Low | Moderate |
CNN/RNN | Captures patterns and sequences, handles text effectively | Limited context range, high overfitting risk | Sequence modeling, text classification | Medium | Moderate | Moderate | High |
BERT/RoBERTa | Deep semantic understanding, context-aware | Ignores network context, computationally intensive | Complex text understanding, sentence-level tasks | High | Low | High | Very High |
GNNs (GCN, GAT) | Models’ relationships, social propagation, network-level insights | Weak on textual semantics | Social network analysis, fake news propagation | High | Low | High | High |
Existing Ensemble Models | Improved performance using hybrid features | Often non-adaptive and dataset-specific | Multi-modal detection, hybrid systems | Medium to High | Low to Moderate | Medium to High | Very High |
