Kansei engineering modelling of intelligent food packaging for beef products using artificial neural network optimisation

Prediction is one of tasks in the application of artificial neural network (ANN). The utilisation of ANN has recently become widespread for predicting product designs, but most research only used one algorithm to train a small dataset. Therefore, this research aimed to predict the design of Kansei engineering-based intelligent food packaging (IFP) for beef products. The dataset comprised 418 inputs, derived from combinations of 19 Kansei words and 22 categories of packaging design attributes. An ANN model was developed and trained by comparing 11 learning function algorithms. This research addressed the gap in predicting the design of IFP using various ANN training algorithms. The results showed that ANN trained with the gradient descent backpropagation algorithm (traingd) provided the highest accuracy. Traingd showed the best fit with the highest R and R2 values as well as the lowest MSE, MAD, and RMSE of 0.9949, 0.98915, 0.0333, 2.1353E-05, and 0.00043656, respectively.