Prediction of Effect of Natural Antioxidant Compounds on Hazelnut Oil Oxidation by Adaptive Neuro‐Fuzzy Inference System and Artificial Neural Network

Abstract

In this study, natural compounds including gallic acid, ellagic acid, quercetin, β-carotene, and retinol were used as antioxidant agents in order to prevent and decrease oxidation in hazelnut oil. Quercetin showed the strongest antioxidative effect among the antioxidative agents, during storage. The accuracy of adaptive neuro-fuzzy inference system (ANFIS) and artificial neural network (ANN) models was studied to estimate the oil samples' peroxide value (PV), free fatty acid (FFA), and iodine values (IV). The root mean square error (RMSE), mean absolute error (MAE), and determination coefficient (R(2)) statistics were used to evaluate the models' accuracy. Comparison of the models showed that the ANFIS model performed better than the ANN and multiple linear regressions (MLR) models for estimating the PV, FFA, and IV. The values of R(2) and RMSE were found to be 0.9966 and 2.51, 0.6269 and 88.55, 0.5120 and 101.8 for the ANFIS, ANN, and MLR models for PV in testing period, respectively. The MLR was found to be insufficient for estimating various properties of the oil samples.

Prediction of oxidation parameters of purified Kilka fish oil including gallic acid and methyl gallate by adaptive neuro‐fuzzy inference system (ANFIS) and artificial neural network

“…The results indicated that the ANFIS model with a high coefficient of determination ( R 2 = 0.999) had better performance compared to ANN ( R 2 = 0.89). Moreover, the ANFIS used was a more effective modelling technique for the estimation of PV of hazelnut oil in the presence of natural antioxidant compounds during storage in comparison with ANN . In another study, the rheological characteristics of different molasses types were satisfactorily predicted in the ANFIS model with high coefficient of determination ( R 2 = 0.979–0.999) and low root mean square error (RSME) (0.12–0.46), depending on the model structure such as membership function type and number .…”

“…Processing the information occurs in each neuron at the same time as other units do. Indeed, the network is trained with a dataset of observations (inputs and outputs) and optimised based on its efficiency to predict a set of known outcomes . As is known, food structures are non‐linear systems which ANNs can successfully model their parameters.…”

Evaluating the potential of artificial neural network and neuro‐fuzzy techniques for estimating antioxidant activity and anthocyanin content of sweet cherry during ripening by using image processing

“…Humans live in a highly oxidative environment, and many processes involved in metabolism may result in the production of excess free radicals, which can lead to health problems . Oxidation causes the formation of toxic compounds that are harmful to human health, such as reactive oxygen species and free radicals, which can lead to carcinogenesis, mutagenesis, inflammation, DNA changes, aging, cardiovascular disease and nutritional loss . In addition, oxidation may result in unpleasant flavor and rancid taste .…”

A fuzzy-wavelet neural network model for the detection of meat spoilage using an electronic nose

“…In addition, CFWNN's prediction results are compared with those obtained by MLP networks and adaptive neuro-fuzzy inference system (ANFIS) identification models. Such schemes have become popular modeling techniques in food science and technology in recent years [14]. In the proposed CFWNN, 10 final rules have been created, using the clustering pre-processing stage.…”