Financial Foresight: Predictive Power of CNN-SVM in Fake Currency Detection Analysis

The stability of national economies depends on the resilience of their currency verification. This paper conducts a comprehensive analysis of counterfeit detection using the CNN-SVM combined approach. The study classifies the detection into five categories: Optical Analysis, Physical Property Analysis, Chemical Analysis, Sensor-Based Detection and Digital Image Processing. The CNN-SVM model was thoroughly trained and tested using a 6500-picture dataset that included actual cash and counterfeit notes. Consequently, it attained an excellent overall accuracy of 97.2071%. For every class, the accuracy recalls and F1-score of the model were calculated with precision. Five classes were assessed for precision; Class 1 yielded a precision value of 95.48%, while Class 2 attained an estimate levelling on exactly ninety-six and point thirty-seven per cent (%). Furthermore, results obtained in dealing with three to five resulted in seventy-nine per cent (%), eighty per cent (% ), and one hundred. The recall rates also showed a similar trend in Class 3, with a commendable score of 99.14%. The F1 scores, indicative of the model's balanced performance, achieved a maximum value for Class 3, reaching as high as 98.27%, demonstrating its distinctive nature in determining difficult counterfeits from others. The model's performance in different classes was evaluated using macro, micro and weighted averages. So, the macro averages of accuracy recall and F1-score were calculated to be 97.13%, 97.27%, and 96/5, respectively. Specifically, the weighted mean micro averages generally reached 97.21% for everything measured. The results indicate the remarkable ability of CNN-SVM system in detecting fake money, setting a new trend for AI use in