Determining the Geographical Authenticity of Sri Lankan Regional Teas with Their Chemical Profile Using Artificial Neural Networks

Abstract

Characterization of authenticity is a timely requirement in food control and safety. The geographically authenticated food products always realize high market prices and sustainable market share. The present study was conducted to investigate the potential of chemical profile in authenticating geographical differences of Sri Lankan regional black teas using artificial neural networks. Thirteen different geographical origins under eight different basses were evaluated and tea samples were chemically analyzed in monthly intervals for a period of one year. Training of feed-forward back propagation multilayer perceptrons resulted in accuracy rates ranging from 46% - 82%. Out of eleven chemical parameters considered for the study, Thea Rubigin and Total Colour are the most important discriminators of geographical origins. Other than that, Gereniol, Thea Flavin, Trans – 2 – Hexanol, Total Polyphenol and Amino Acids are also potential discriminators. The study concluded that the differentiation and classification of Geographical Origins (GOs) of Sri Lankan regional black teas is possible using the profile of chemical attributes and applying artificial neural networks. Finally, the data obtained demonstrates that it is possible to assign unknown samples according to their region of origin. Further development of this work could lead to a simple interface that could be used by brokers or exporters to identify the geographical origin of an unknown tea sample.