Categorizing Functional Yoghurt using Artificial Neural Network

Yoghurt was supplemented with low molecular weight carbohydrates (LMWC) extracted from Syzygium cumini seeds. Total soluble solids, pH, color, titratable acidity, texture, sensory and shelf life studies were quantified in control and functional- F1 (1% LMWC) and F2 (5% LMWC) yoghurts over a period of 15 days. An artificial neural network (ANN) was developed that could classify the yoghurts with color, pH and % carbohydrate as inputs. The ANN with one hidden layer in a feed forward pyramidal framework was trained using the gradient descent algorithm to reach an MSE (Mean of Squared Errors) of 0.055314. Of the total 120 data points, 30, 60 and 30 were randomly chosen for training, testing and prediction. The ANN could classify the yoghurts with 100% efficiency (r = 0.95). This study presented a minimally invasive approach that can classify functional food products on the basis of physical and chemical properties to determine user acceptability.