Fraud Detection Using Machine Learning in Blockchain Stock Marketing Transactions

One of the main concerns relevant to blockchain-based stock market transactions is fraud detection, as it is vital to protect market integrity and investors‘ trust. The existing systems have many limitations, as these are usually focused on rule-based approaches and, as a result, false positives and false negatives are common. The study in question shows that a novel machine learning approach helps to address the problem. The proposed system utilizes various supervised, unsupervised, and ensemble learning techniques to ‘learn’ to evaluate the transactional data inside the blockchain networks. The methodology helps to ensure that detection fraud will be highly efficient, and it involves data preprocessing, algorithms selection for training, model tuning, and implementing real-time monitoring. The results indicate that the novel system outperforms many existing ones, as its accuracy is 92%, precision is 91%, recall is 94%, and F1-score is 0.92. Moreover, false positives and false negatives account for 5% and 6%, which is significantly better compared to existing solutions. As a result, fraud detection in the studied context is strongly impacted by the proposed solution.