Exploration of Fast Edible Oil Classification Using Infrared Spectrum, Machine Learning, and Chemometrics

Your food is your medicine. Edible oils take important parts in people's daily food, and taking good-quality oils plays an important role to the health. However, as the supply cannot satisfy the market demand and some good-quality edible oils are expensive, many incidents of adulterated and fraudulent edible oils have been reported. In Taiwan, some common adulterated edible oils and fraudulent edible oils incidents include (1) mixing good-quality oils with low-quality oils, but labeling the products as high-quality products; (2) importing cheap and low-quality oils abroad, but labeling them as good-quality ones; and (3) fraudulent labeling with wrong ingredients. Even though high-tech laboratories can differentiate the products and identify ingredients, the popular technologies demand high costs in terms of money, time, and man power. The general cannot easily access these technologies and should only depend on occasional reports from the governments or from some trusted institutions. Furthermore, the jurisdiction process takes a long time, and the punishment is relatively light, compared to the illegal interests. It is, therefore, crucial to develop new technologies that can effectively and efficiently differentiate different edible oils or even identifying concerned ingredients in edible oils. Due to dropping prices of infrared spectroradiometers and advances in machine technologies and chemometrics, we would like to integrate these technologies to develop a process that can fast and effectively differentiate different edible oils and even identify suspicious ones. The preliminary experiments show some promising results and potential. We also point out some challenges for future work.