Evaluating Machine Learning Algorithms for Fake Currency Detection

Abstract

Currency is a critical asset in any economy, yet it is vulnerable to counterfeiting, undermining its value and disrupting economic stability. Counterfeit currency is particularly prevalent during economic transition, such as demonetization, as fake notes are circulated to mimic real currency. Due to the subtle similarities between genuine and fake notes, distinguishing between them can be challenging. Consequently, financial institutions like banks and ATMs require robust automated systems to accurately detect counterfeit currency. In this study, we evaluate the effectiveness of six supervised machine learning algorithms—K-Nearest Neighbor, Decision Trees, Support Vector Machine, Random Forests, Logistic Regression, and Naive Bayes—in detecting the authenticity of banknotes. Additionally, we examine the performance of LightGBM, a gradientboosting algorithm, in comparison to these traditional methods. Our findings contribute to developing reliable, automated systems for counterfeit detection, and enhancing financial security.