AI and Machine Learning Approaches for Fake Bank Currency Detection: A Detailed Study on Algorithm Efficiency and Scalability

Abstract

The stability of a nation's financial system heavily relies on the legitimacy of its currency. However, fake notes often enter circulation, closely resembling genuine currency, making it difficult for people to differentiate between counterfeit and real notes, despite the presence of various security features. This issue was particularly evident during the demonetization phase, when a significant amount of fake currency emerged. To tackle this problem, it is imperative to have an automated system in place for detecting counterfeit banknotes in banks and ATMs. This study delves into the development of such a system by utilizing supervised machine learning algorithms on a datasets from the UCI data world Machine Learning Repository to identify counterfeit currency. The algorithms utilized include Support Vector Machine, Random Forest, Decision Tree, and K-Nearest Neighbor. Each model underwent training and testing using three different train-test splits: 80:20. The performance was assessed based on metrics such as Precision, Accuracy, Recall, F1-Score, and MCC. Certain algorithms exhibited 95% accuracy under specific train-test ratios, indicating their potential for dependable currency authentication.