ARTIFICIAL NEURAL NETWORK (ANN) MODELLING USING DIFFERENT ALGORITHMS FOR PREDICTING THE TOTAL PHENOLIC CONTAIN IN COCOA SHELL EXTRACT.
Authors: Md Yusof A. H. , Abd Gani S. S., Zaidan U. H., Halmi M. I. E.
Journal Issue: Malaysian Cocoa Journal, Volume 14/2022
Keywords: Artificial neural network, cocoa shell, phenolic content
Published On: 10/05/2022
Abstract
In this study, three algorithm networks, Increment Back Propagation (IBP), Batch Back Propagation (BBP), and Genetic Algorithm (GA) for ANN model were used to predict the total phenolic content from cocoa shell extract. The data were divided into two sections, 75% training set and 15% the test set. All algorithms were training, testing, and calculating its Root Mean Square Error (RMSE), Average Absolute Deviation (AAD), and Coefficient of Determination (R2). The model result was compared and show a high correlation of coefficient (R2 IBP 0.9997, R2 BBP 0.9997, R2 GA 0.9996). This showed the ANN using three different network algorithms was able to predict the total phenolic content. The ANN model with BBP shows better prediction data with a higher R2 value and smaller RMSE (IBP 0.3131, BBP 0.0622, GA 0.3068) and ADD (IBP 1.1987, BBP 0.2989, GA 1.2048). This finding suggests that the ANN with algorithm BBP showed a better prediction and fitting ability compared to the BBP and GA.