Modeling of sunflower oil treated with lemon balm ( Melissa officinalis ): Artificial neural networks versus multiple linear regression

Abstract

This study aimed to develop, evaluate, and compare the performance of artificial neural networks and multiple linear regression models in the estimation of phenolic profile of sunflower oil enriched by lemon balm. Total phenolic material in addition to the quality parameters (induction time and antioxidant activity) of the treated oil was compared to those of the pure sunflower oil. The oxidative stability of the product was increased by almost 7% in terms of induction time, while the phenolic profile was increased by almost 2.5 times. Moreover, the antioxidant activity of sunflower oil was enhanced by ~5 times over the pure oil. The values of artificial neural networks and multiple linear regression were calculated as: error rates 0.01% and 8.09%; root-mean-square error values 0.45, and 4.36; R2 values 0.9958 and 0.6183, respectively. Practical applications Although sunflower oil is one of the most consumed food products worldwide, its susceptibility to deterioration due to its low oxidative stability is major obstacle in the related industry. Therefore, it becomes inevitable to develop the product as a functional food both in order to improve the product in terms of health, and to prevent undesirable situations by extending its shelf life.