Application of artificial neural network to simulate phenolic content and antioxidant activity during spontaneous fermentation of Garcinia Mangostana pericarp / Muhammad Musyedi Mazri and Mohamad Sufian So'aib

• Main Authors: Mazri, Muhammad Musyedi (Author), So'aib, Mohamad Sufian (Author)
• Format: Book
• Published: 2020.
• Subjects: Conference or Workshop Item; PeerReviewed
• Online Access: Link Metadata
• Notes: https://ir.uitm.edu.my/id/eprint/81539/1/81539.pdf
• Keywords: Botanical chemistry. Phytochemicals; Nutrition. Plant food. Assimilation of nitrogen, etc.

In spontaneous fermentation process and antioxidant activity in fruits, there is unpredictable nature of the spontaneous fermentation that cannot be able to predict. While Artificial Neural Network (ANN) is a model for the signal processing modelling which is appropriate for prediction to solve the problem like classification problems and predictions. The ability of the artificial neural network can do as well as human brain such as learn a new thing and adapt to the new changing of environment. ANN architecture modelling is required to solve this critical problem which is the prediction of non-linear pattern of the activities. The network is consisting of three layers which is input layer, the hidden layer and the output layer. The method use in this modelling is Levenberg-Marquardt backpropagation training function of neural network since the method is the simplest among the other artificial neural network modelling. Several trials were made by using different of transfer function which is "tansig", "logsig" and "purelin". The ANN model used NN 2-7-1 neurons in input-hidden-output layers. The model developed which is NN 2-7-1 has an acceptable generalization accuracy and capability. The predictive ability of the ANN methods by assessed the basic of the mean square error (MSE), and coefficient of determination (R2) between the predicted values of the networks and the actual result from experimental data. the efficiencies of ANN modelling can be concluded by observed the result of MSE, and R2. The minimum value of mean squared error (MSE) and the regression value (R-value) which is closed to 1 showed that the neural network architecture was performed with high accuracies. For total phenolic content, R= 0.99157 while for antioxidant activity, R= 0.99879 respectively. Mean squared error (MSE) showed a very good result from ANN model which is for phenolic content testing value was equal to 0.0009697 while for antioxidant activity testing value was equal to 6.89e-05. As a result, ANN modelling was effectively simulated and predicted the total phenolic content and antioxidant activity in Garcinia Mangostana pericarps.