Perancangan Sistem Pembayaran Non-Tunai Menggunakan Face Recognition sebagai Metode Pengenalan Identitas dengan Algoritma Convolutional Neural Network

Abstract

The development of technology brings us into a very fast acceleration of science, so that the old technology that is still used such as Payment Instrument Using Card already has more and more loopholes. Thus, reported violations of the use of card are also increasing. On the other hand, the rapid development of technology in the era of the Industrial Revolution 4.0, has had many positive impacts on humanity so that it has been widely used in various fields. One of the most famous examples is Artificial Intelligence (Al) which is utilised for the development of science and technology. As in China, many have used face recognition as a medium for identity recognition, such as login, registration, attendance and others. In Indonesia itself, PT. KAI (Kereta Api Indonesia) has used face recognition technology in the boarding process. In this study, the author designs a payment system with face recognition as identity recognition using the Convolutional Neural Network (CNN) algorithm. This algorithm itself has been widely used as a learning algorithm with a high level of accuracy, but a lot of learning data is required. Face image data used in model training is 6000 images in total divided into 10 classes or labels. For testing, 15% of the dataset is selected, namely 900 randomly selected testing data. The model used uses 7 layers of convolution and 3 layers of neural network and activated using softmax function. This design also produces a product purchase web-based application designed using the PHP programming language, with the help of MySQL DBMS. Furthermore, the display is decorated using templates from the Bootstrap 4.6 framework. The results of this design produced a model with an average accuracy of 99% and an average precision of 97%. There is nothing that indicates that the resulting model is overfitting because the accuracy curve and training loss are good.