Detection of Counterfeit Currency Using Supervised Machine Learning Algorithms
Counterfeit currency is an extremely common yet pertinent problem of presence of fake, inauthentic or copies of real currency in the market and economy that is faced by various nations across the globe. Cash transactions still make up for over 80% of all transactions. Fake and inauthentic notes therefore continue to be a major source of nuisance for the economy. In this paper, the aim is to identify the authenticity of the currency notes by using various machine learning algorithms and also to compare and contrast which of these algorithms is best suited for the same. The machine learning algorithms classify the currency notes on the basis of features extracted from images. This paper involves a two-part study; in the first part, the dataset was taken from UCI Machine Learning Repository. We got the best results from K-Nearest Neighbors Classifier with 99.8% accuracy and F-Score of 0.992 (β=2). In the second part of study, the dataset was generated from images on Indian banknotes, real as well as fake and the best results were derived from Gradient Boosting Classifier with 99.9% accuracy and F-Score of 0.998.
Keywords–Counterfeit Currency, Discrete Wavelet Transform, Machine Learning, K-Nearest Neighbors, Supervised Learning, Gradient Boosting Classifier