Neural Network Modelling of Phenolic Content, Antioxidant Capacity and Microbial Population Dynamics of a Household Scale Spontaneous Fermentation of Carica Papaya Leaf

• Published in: Journal of Mechanical Engineering
• Main Author: Latip N.A.; So’aib M.S.; Tan H.L.; Senin S.F.; Hamid A.
• Format: Article
• Language: English
• Published: UiTM Press 2022
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85147765002&doi=10.24191%2fjmeche.v11i1.23614&partnerID=40&md5=c97fcb8e0088df710c5e4e1ee844c163

Beneficial effects of spontaneous fermentation on Carica papaya leaf (CPL) have been observed in terms of enhanced total phenolic content and antioxidant capacity, as well as cultivation of lactic acid bacteria (LAB). Nonetheless, these responses were nonlinear, thus Artificial Neural network (ANN) was used as a predictive tool. The chosen ANN architecture consisted of multi-layer perceptron (MLP) with 2-7-7-1 and 2-10-10-1 topologies,Levenberg-Marquardt training algorithm, and hyperbolic tangent sigmoid activation function. Enhanced total phenolic content (TPC) and antioxidant capacity were recorded in final CPL extracts (day 90) of 5-L fermenter origin; 48.42±0.31 mg GAE/g dm and 467.38±4.09 mM TE/g dm, respectively, as compared to 12.13±0.39 mg GAE/g dm and 275.46±3.09 dm of respective extracts at initial (day 0). Likewise, enhanced total phenolic content (TPC) and antioxidant capacity were also observed for 50-L fermenter origin extracts. The chosen ANN topologies displayed the highest predictive ability as indicated by their correlation coefficient (R) of greater than 0.9, the marginal difference in mean square error (MSE) between training and testing data sets, and the absolute average deviation (AAD) of less than 10% between the predicted and experimental values of most responses. In conclusion, ANN was a reliable predictive tool for nonlinear responses during spontaneous fermentation of CPL. © 2022 College of Engineering, Universiti Teknologi MARA (UiTM), Malaysia.