Maximizing the Accuracy of Fake Indian Currency Prediction Using Particle Swarm Optimization Classifier in Comparison with Lasso Regression
R. Kishore Kumar, C. Nelson Kennedy Babu, A. Akilandeswari · 2025
This research aims to assess and compare the effectiveness of two machine learning (ML) techniques in detecting counterfeit currency. Specifically, it employs neural networks through the use of particle swarm optimization (PSO) and Lasso regression (LR) classifiers. The research methodology includes collecting a dataset, which is then divided such that 80% is utilized for training the proposed PSO model and the remaining 20% for testing. Outputs from both classifiers are organized into two sets, each containing 10 values from various operations, totaling 20 for SPSS statistical analysis. This analysis employs a 95% confidence interval (CI) and a G power of 0.95. The dataset comprises 1,050 variables related to currency features like size, color, and unique markers. After a preliminary examination, the performance of the PSO and LR classifiers in accurately identifying counterfeit notes is compared. Results indicate that the novel PSO classifier exhibits superior accuracy in distinguishing fake currency from genuine notes compared to the LR classifier. However, statistical analysis reveals no significant difference between the novel PSO and LR algorithms, with a p-value of 0.436 (p>0.05), suggesting that while the novel PSO has higher accuracy, this difference is not statistically significant when compared to LR's performance.