Predicting Peroxide Value of Peanut Oil using Machine Learning Models

Citation: Chen, YT, Ling, MHT, Yang, H, Cai, MH, Koh, RW, Tan, RYH, Xue, X. 2024. Predicting Peroxide Value of Peanut Oil using Machine Learning Models. Acta Scientific Nutritional Health 8(7): 116-122.

Link to [abstract].

Here is the permanent [PDF] and [Dataset] links to my archive.

Natural antioxidants (NATOs) derived from sources like rosemary, green tea, and oregano have acquired extensive attention for their efficacy in preserving edible oils, presenting a promising alternative to synthetic antioxidants (ATOs) due to their superior safety profile. However, integrating NATOs into the food industry faces challenges stemming from the variability in their chemical composition, necessitating prolonged stability tests based on peroxide values (PV). This study explores the predictability of PV in peanut oil using three chemical parameters (total phenolic content, total antioxidant content, and total carotenoid content), one physical parameter (partition coefficient), and storage duration. Six machine learning classifiers (logistic regression, multilayer perceptron, radial basis function, Gaussian Naïve Bayes classifier, support vector machine, and decision tree) were employed. The results have shown significant correlations between the chemical parameters and antioxidant activity. Our findings indicate that PV in peanut oil can be accurately predicted using these parameters and storage duration, with the multilayer perceptron demonstrating the highest predictive performance, achieving an accuracy of at least 89.8% in determining whether PV remains within acceptable limits post-storage.