Accurate Detection of Forgery in Signature using Deep Learning Algorithm in Comparison with Random Forest Model
P. Jayaprakash, G. Ramkumar · 2023
This research is about the detection of forgery in signature images from the parameters taken from the data set using the Novel Random Forest Model (RFM) deep learning algorithm to compare the accuracy with Convolutional Neural Network-xg (CNN-xg). In total, 44 samples are taken for this study, with those samples being split evenly between two groups of 22, making the total number of samples gathered 44. Group 1 makes use of something called the Random Forest Model (RF), whereas Group 2 makes use of something called CNN-xg. In order to facilitate the creation of the code and its subsequent deployment, the machine learning capabilities of the Colab software were leveraged. For the purpose of determining the appropriate value for the sample size, the sample size is determined with the assistance of an online statistical analysis tool, a power value of 80%, and an alpha value of 0.05. This is done in order to arrive at the correct value for the sample size. The data that were obtained from the earlier studies are used. From simulation results, the Convolutional Neural network-xg Algorithm obtained a prediction accuracy of 97.43% and Random Forest Model (RF) a predicted accuracy of 72.291% with significance values of 0.040 (p<0.05). For the given dataset Convolutional Neural Network-xg performs significantly better than the RF in finding the accuracy for detecting forgery in signature images.