Development of a Machine Learning-Based Fake News Detection Model for Nigerian News Content

Fake news poses a significant challenge in today's digital landscape, particularly in Nigeria, where misinformation can lead to social unrest and erosion of trust in the media. This study presents the development of a machine learning-based system specifically designed to detect fake news in Nigerian news content. A dataset was created by scraping real news articles from reputable Nigerian news websites like Punch, Vanguard, and Legit.ng, and generating fake news articles with the assistance of local journalists. The fake news articles were crafted to reflect common characteristics such as sensational language, exaggerated claims, and emotional triggers.
The system employs natural language processing techniques, such as TF-IDF vectorization, alongside machine learning algorithms, including Logistic Regression, SVM, Decision Tree, Gradient Boosting, and Random Forest. An ensemble model integrating these classifiers achieved high accuracy (93%) in distinguishing fake news from real news. The system was further enhanced with semantic similarity checks, fake news indicators, and a user-friendly GUI for manual testing.
Evaluation results show the effectiveness of ensemble learning in achieving robust predictions, with Random Forest and Ensemble model (voting classifier) demonstrating the highest accuracy. The study highlights the importance of tailoring detection systems to the linguistic and cultural context of Nigerian media, providing a valuable resource for combating fake news and fostering an informed society.