Enhancing Fake News Detection Using BERT: A Comparative Analysis of Logistic Regression, RFC, LSTM and BERT
Bibliographic record
Abstract
The research analyzed how to identify fake news effectively, understanding the developing capabilities to differentiate between misinformation and the correct nature of news in today’s complex media environment. Numerous approaches were tested: mostly the ways of determining if a certain message is valid or false. With application of Logistic Regression, Random Forest and lastly LSTM models, the research has been able to achieve 98.7%, 98.8%, and 95% accuracies respectively, showing that these traditional and advanced techniques are very effective. However, it became apparent that these models performed poorly as soon as tested on a new dataset and the discrepancy was not being able to understand context while being trained. Understanding the context is the key in news identification, so the research turned to BERT, the Transformer model pre-trained with huge contextual data to perform this task. Drawing on its deep and diverse knowledge base, BERT has shown a remarkable aptitude for sorting news articles into fake/real categories based on their context. Logistic Regression, Random Forest, and LSTM models demonstrated that although they were able to build models that were highly accurate up to 99% on familiar data, their accuracy dropped disproportionately as soon as new data was given. BERT, despite having a lower overall accuracy of 84% demonstrated a better sensitivity to contextual nuances in the news data. Here, the importance of contextualized conception in the sphere of fake news detection should be emphasized as a way to take advantage of BERT’s complicated comprehension which is a promising alternative of more accurate and effective identification as the media landscapes of the modern times may be complex and diverse. Despite the fact that the classical models surpass in planned environments, BERT’s capability of contextualization made it irreplaceable for evidence as the news sources in uncommon circumstances can digress from base.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.011 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.003 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".