Deep-Neural Genetic Algorithm Optimization Analysis for Forgery Detection of Banknotes
Sashank Sridhar, Sowmya Sanagavarapu · 2021
Metaheuristic algorithms aim to find high performing, near-optimal solutions with reasonable computing costs. Using the process of natural selection that belongs to a set of evolutionary algorithms, are Genetic Algorithms. These algorithms intelligently exploit random search over the data in search of a solution space for better performance. Detection of forgery in legal documents by an automated system remains a valid problem of today. In this paper, the identification and detection of forgery in the monetary notes is performed by using deep neural-based Genetic Algorithms. An Artificial Neural Network was used for training on the banknote dataset and the learned weights were vectorized for Genetic Algorithm processing. Genetic Algorithm is explained in detail along with the working of the Fitness model that determines the efficiency of the trained model. The model achieved a high-performance accuracy of 94% with the fitness score of95.2. The improvement of the model with the infusion of trainable weights from a deep neural model is represented and analyzed.