M.B. Anargiansyah, N.N. Laila, F.R.A. Pramudya, B.R.P. Fayensi, S.A. Khairunnisa, N. Fadhilah, M. Khoiro
In Indonesia, many people have resorted to reusing discarded waste cooking oil. Waste cooking oil undergoes decomposition reactions that degrade quality and adversely affect health. In this study, RNNs (Recurrent Neural Networks) algorithms are employed to determine the quality of the cooking oil. Quality palm cooking oil is classified into three classes based on color: new (yellow), decent (yellow orange), and undecent (blacked-brown). RNNs can provide consistent and highly accurate identification results with fast processing times. Conventional observations are only sometimes consistent in assessing color or oil quality, especially when repeated assessments require a long time. RNNs allow for objective identification of oil quality, eliminating human subjectivity. The urgency of this research is to offer a new technology that simplifies identifying oil quality quickly, inexpensively, and easily accessible to the public. The RNN model employs an extensive database of image data, enabling more accurate identification than conventional observation methods. The data were collected from three various cooking oil types for each class. These samples are photographed using four smartphones under different lighting conditions, resulting in a total of 3,600 data points. The RNNs classification model uses 2,700 training data and 900 test data. Based on the research, the classification of cooking oil quality using the RNNs model achieves an accuracy rate of 70.33%. However, the model exhibits some misclassification errors, particularly in categorizing suitable oil unsuitable. Nevertheless, the developed RNNs model yields promising results overall. © 2026 Author(s).
Physics Department, Faculty of Mathematics and Natural Sciences, Universitas Negeri Surabaya, Surabaya, Indonesia
