Evaluating Supervised Learning Techniques For Accurate Fake News Identification Using Support Vector Machine Technique

Abstract

In this era of technological advancement where the internet has revolutionised communication, users generate and share vast amounts of information, some of which is misleading and not real. Automated identification of misinformation or disinformation in textual articles poses a significant challenge. This study focused on distinguishing between fake and true news using a Support Vector Machine (SVM) Classifier trained on Term Frequency-Inverse Document Frequency (TF-IDF) vectorization method. The dataset was split into 80% for training and 20% for evaluation. Model performance was evaluated using a Confusion Matrix, Classification Report, and ROC curve. The Confusion Matrix depicted accurate predictions of 6450 fake news articles correctly classified and 215 true news misclassified as fake. Similarly, 7347 true news articles were correctly identified along with 86 fake news misclassified as true. The overall model achieved an accuracy of 97.86%Precision scores were 98.68% for fake news and 97.16% for true news, with recall scores of 96.77% for fake news and 98.84% for true news. fake news category attained an F1-score of 0.9772 while true news events attained an F1-score of 0.9799. The ROC curve, with an AUC value of 1.00, demonstrated the model's excellent diagnostic ability in distinguishing between fake and true news articles.