Fake news refers to deliberately false or misleading information presented as factual news. The rise of fake news has led to the need for efficient detection techniques to prevent its harmful effects on individuals and society. Previous fake news detection techniques relied on features such as linguistic cues and stylistic features. However, these techniques had limitations in terms of accuracy and generalizability. To overcome these limitations, this paper presents a machine learning algorithm for detecting fake news namely an enhanced stacking ensemble classification algorithm (ES-ECA). The proposed algorithm uses preprocessing techniques to split the dataset into individual statements and generates n-grams as features. Feature extraction is then performed using the term frequency-inverse document frequency (TF-IDF) measure, which quantifies the importance of each feature to the document. The proposed algorithm uses an enhanced stacking ensemble approach that combines base classifiers, including an enhanced version of the J48 decision tree algorithm, an enhanced version of the Naive Bayes algorithm, and an enhanced version of the k-Nearest Neighbors algorithm. A meta-classifier is then created using a Random Forest algorithm boosted using the AdaBoostM1 algorithm. The RandomSubSpace algorithm is used to improve the performance of the stacking ensemble classifier. Experimental results show that the proposed algorithm outperforms previous techniques in terms of accuracy, precision, recall, and F1-score. The ensemble classifier achieved an accuracy of 75.18% and an F1-score of 81.81%, outperforming the individual classifiers. These results suggest that the proposed algorithm is effective in detecting fake news and can be used to mitigate the harmful effects of fake news on society.

Classification, stacking, preprocessing, n-gram, TF-IDF