A Study on Fake News Detection Using Machine Learning and Deep Learning

RUPALI SANGARE , ABHISHEK KUMAR , DR. HARSHALI PATIL , DR. JYOTSHNA DONGARDIVE

Fake News Detection, Machine Learning, NLP Techniques, Text Analysis, Scikit-learn, Feature Selection, Precision Evaluation, Media Platforms, Misinformation, and Supervised Learning.

The proliferation of misinformation in the digital age poses a complex challenge, impacting public opinion, elections, and societal harmony. This study focuses on effective fake news detection, employing supervised machine learning algorithms as a solution. A comprehensive literature review of fake news detection methodologies precedes the proposed approach. This exploration encompasses linguistic analysis, source credibility assessment, and computational models, establishing the foundation for an optimized detection framework. The methodology revolves around implementing supervised machine learning, leveraging Python's scikit-learn and NLP techniques for text analysis. Precise feature extraction is pivotal, enabling accurate differentiation between genuine and deceptive information. Utilizing tools like Count Vectorizer and Tfidf Vectorizer in scikit-learn converts textual data into numerical feature vectors, facilitating model training. Feature selection assumes significance in enhancing classification precision. Employing techniques like chi-square analysis, information gain, and recursive feature elimination aims to prioritize influential features for accurate fake news classification. The evaluation metrics, particularly precision derived from confusion matrix outcomes, guide optimal feature selection for the final model. This research encompasses extensive experimentation, systematically refining the model's performance through cross-validation. Iterative processes involve hyper parameter tuning, diverse algorithm exploration, and assessing robustness across datasets. Emphasizing precision minimizes false positives, crucial for curbing fake news dissemination while safeguarding genuine information. Insights gained from this study hold implications for combating misinformation, informing policy interventions, and designing more effective detection mechanisms in digital ecosystems.

Volume 9 Issue 1, January-2024

Pages : 551 - 560