Deteksi Bahan Makanan Non Halal Berbasis Machine Learning dan Cloud Computing

Abstract

Indonesia has a significant Muslim population. In Islamic teachings, consuming halal food is an obligation for Muslims. However, in the current era of globalization, Muslims need to maintain the halalness of the products they consume, especially people who are visiting or are in countries that have a majority non-Muslim population. Several previous studies have successfully created a system or application to detect the halal status of a packaged food or beverage product. In this research, the author proposes the development of the technology used, especially for imported food products and can be used by tourists traveling to countries that have a minority Muslim population. This system utilizes OCR technology, Machine Learning and Cloud Computing, and the system is made to support several languages. In this research, the authors implemented Cloud Computing technology and utilized Long Short-Term Memory (LSTM) architecture. The model is trained with a pre-processed composition text dataset, then with 2 LSTM layers, 2 Dropout layers, and 2 Dense layers. The results of the model get a training accuracy of 99.99% and a training loss of 0.00041 and accuracy using real food sample test data of 84% with recall and precision of 88.89%, and F1 Score of 88.90%. This model is then deployed to cloud computing services to be implemented into an android application. It is hoped that this research can provide important benefits for users in assessing the halal composition of packaged food and beverage products, especially for Muslims who are abroad.