Kurdish Fake News Detection Based on Machine Learning Approaches

The widespread use of social media platforms and the internet has increased information sharing, including both true false news. Detecting fake news is challenging, several studies have been conducted to automate this process for popular languages such as English Arabic. However, more research must be done on detecting in low-resource Kurdish. This gap was addressed, a publicly available Kurdish dataset (KDFND) used, comprising 100962 articles, among which 50751 are real, 50211 labeled Real Fake. In study, three techniques were implemented extract features from texts, word embedding, term frequency-inverse document frequency, count vector, various machine learning deep classifiers used (Random Forest, Support Vector Machine, Convolutional Neural Networks) identify dataset. results showed that with textual content could identified, especially when convolutional neural networks used. According experimental CNN performs better than other models, an F1-score 95% accuracy 91% percent. These findings indicate methods can efficiently detect like Kurdish, even complex environments.