FinTechFake: A Linguistic-Enabled Artificial Intelligence Approach for Detecting Fake News in Finance

The spread of misleading information on social media and internet forums has posed a serious danger to organizations and enterprises. This paper proposes an ensemble learning-based artificial intelligence approach called FinTechFake for financial fake news detection. Due to the public unavailability of the financial fake news dataset, we created a financial fake news dataset from the existing benchmark fake news dataset using topic modeling approaches. Substantial feature engineering has been investigated to extract the best-suited linguistic aspects that include sentence cohesion, stance, sentiment, and other grammatical features. Furthermore, linguistic features are combined with word embedding-based characteristics to apply an ensemble learning approach for the automatic detection of financial fake news from large text corpora. In terms of classification accuracy, both XGBoost and LGBM produce good results with 96.8% accuracy. When compared with the existing state-of-the-art, FinTechFake , outperformed others in terms of F1 score by 2.6%.