Özet:

Abstract Social media platforms' utilization for news consumption is steadily growing due to their accessibility, affordability, appeal, and ability to propagate misinformation. False information, whether intentionally or unintentionally created, is being disseminated across the internet. Certain individuals spread inaccurate information on social media to gain attention, financial benefits, or political advantage. This has a detrimental impact on a substantial portion of society that is heavily influenced by technology. It is imperative for us to develop better discernment in distinguishing between fake and genuine news. In this research paper, we present an ensemble approach for detecting fake news by using TF-IDF Weighted Vector with Word2Vec. The extracted features capture specific textual characteristics, which are converted into numerical representations for training the models and balanced dataset with the Random over Sampling technique. The implementation of our proposed framework utilized the ensemble approach with majority voting which combines 2 machine learning models like Random Forest and Decision Tree. The proposed strategy was adopted empirically evaluated against contemporary techniques and basic classifiers, including Gaussian Naïve Bayes, Logistic Regression, Multilayer Perceptron, and XGBoost Classifier. The effectiveness of our approach is validated through the evaluation of the accuracy, F1-Score, Precision, Recall, and Auc curve, yielding an impressive accuracy score of 94.24% on the FakeNewsNet dataset.

Anahtar Kelimeler: