This paper presents a machine learning method for identifying ten different plant oil types and their mixtures based on their fatty acid profiles. The method uses an unsupervised model to identify sub-clusters within oil types, followed by a supervised deep learning model trained on simulated oil mixtures. The model achieves a 50th percentile absolute error of 1.4–1.8% and a 90th percentile error of 4–5.4% for three-way mixtures. An online-training method allows for continuous improvement and adaptation to new oil profiles.

Kevin Lim, Kun Pan, Zhe Yu, Rong Hui Xiao