Comparative Analysis of Machine Learning Algorithms for Fake Currency Detection: A Probabilistic Approach
S. Santhiya, P. Jayadharshini, N. Abinaya, S V Harish, E Navin, N Indrajith · 2024
With the increasing rate of counterfeit currency in the market, the need for efficient fake currency detection systems has become essential. The circulation of fake currency poses a great threat to the stability of financial systems and affects the integrity of economies worldwide. This project presents a comparison on the application of five distinct machine learning algorithms—Logistic Regression, Support Vector Machine (SVM), k-Nearest Neighbors (KNN), Gradient Boost Classifier, and Naive Bayes—in the development of a fake currency detection system. The aim of the this project is to examine the accuracy and other evaluation criteria of these models’ performance in identifying counterfeit money. Leveraging a probabilistic approach, Logistic Regression models the likelihood of a banknote being counterfeit based on extracted visual and texture features from currency images. SVM's ability to operate effectively in high-dimensional spaces makes it particularly suitable for our diverse dataset, accommodating various features crucial for accurate detection. K-Nearest Neighbors (KNN), a proximity-based algorithm, is utilized to classify banknotes by considering the characteristics of their nearest neighbors in the feature space. The Gradient Boost Classifier is employed as it adapts to complex relationships within the dataset, making it well-suited for finding difficult or complicated patterns of counterfeit banknotes. Naive Bayes models the conditional probability of a banknote being counterfeit given its features, offering a computationally efficient solution for real-time detection scenarios. The Gradient Boost Classifier, which obtains the 99% as highest accuracy of the five models, is demonstrated to be effective by the experimental findings.