Fourier-transform infrared spectroscopy and machine learning to predict fatty acid content of nine commercial insects

Insects used as food and feed has attracted considerable attention for their nutritional profiles in recent years. Fatty acid profile, such as unsaturated fatty acid content, ratio of unsaturated to saturated fatty acids, determines the quality of insect products. Multiple previous studies have used spectroscopy technologies and machine learning algorithms to predict fatty acid content in various foods and feeds. However, these approaches were not applied for predicting fatty acid content in insects before. In this study, 50 insect samples containing 9 commercial insect species were collected. Machine learning methods were applied to build the calibration models to predict fatty acid content from Fourier-transform infrared spectroscopy spectra. For all fatty acids, partial least square regression, regression trees and neural network based methods were among the best machine learning methods. For the best performing model, a coefficient of determination of 0.98, a root mean square error of prediction of 3.19%, and a ratio of performance of 3.91 were achieved using regression tree to predict linoleic acid. The high model performance indicates the potential of applying FTIR and such machine learning methods for fast and non-destructive prediction of fatty acid of insect oil products.