Prediction of Trans-anethole Extraction Yield from Pimpinella Anisum Seeds Using Artificial Neural Network

Abstract

In this investigation, the extraction of trans-anethole (t-anethole) using subcritical water solvent was employed as a case-study. A feed-forward
multilayer back propagation artificial neural network (ANN) with various train algorithms and number of neurons was considered for the
prediction of t-anethole extraction yield (mg/g dry sample). The input variables were temperature (100-175 oC), flow rate (0.5-4 ml/min), mean
particle size (0.25-1 mm) and output was t-anethole extraction yield. The optimization of neural network structure is manufactured based on
minimum mean square error (MSE) of training and testing data. The optimal ANN model is composed of one hidden layer with five neurons.
The prediction of t-anethole extraction yield using the ANN model was confirmed to be an accurate, appropriate and simple method.