engleski

Application of linear, nonlinear and artificial neural network modelling for description of the medical plants extraction process

The objectives of this study were to use multiple linear regression (MLR), nonlinear regression (NLR), piecewise linear regression (PLR) and artificial neural network (ANN) modelling to analyse the effect of the extraction time, the extraction temperature and plant species on total dissolved solids, extraction yield, total polyphenols and antioxidant activity of three medical plants aqueous extracts. The aqueous extracts of plants from Lamiaceae family, lavender (Lavandula x hybrida L.), melissa (Melissa officinalis L.) and mint (Mentha L.) were prepared and sampled at regular time intervals (0.5, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 70, 80, 90 min) and analysed for physical and chemical properties. The performances of the proposed MLR, NLR, PLR and ANN models were evaluated based on correlation of determination (R2) and root mean square error (RMSE). The results showed that ANN has higher prediction capability compared to other developed models. The relative importance of the variables on the physical and chemical properties of analysed extracts were also determined by global sensitivity analysis. It was determined that the extraction time showed the highest influence on physical and chemical properties of analysed medical herbs extracts.

linear model, nonlinear model, artificial neural network, medical herb extract

nije evidentirano