Ultrasound assisted extraction of phenolic compounds from P. lentiscus L. leaves: Comparative study of artificial neural network (ANN) versus degree of experiment for prediction ability of phenolic compounds recovery

Abstract Design of experiments (DOE) based on central composite design (CCD) and artificial neural networks (ANNs) were efficaciously applied for the study of the operating parameters of ultrasound assisted extraction (UAE) in the recovery of phenolic compounds from P. lentiscus leaves. These models were used to evaluate the effects of process variables and their interaction toward the attainment of their optimum conditions. Under the optimal conditions (13.79 min extraction time, 33.82 % amplitude and 30.99 % ethanol proportion), DOE and ANN models predicted a maximum response of 140.55 and 138.3452 mgGAE/gdw, respectively. A mean value of 142.76 ± 19.98 mgGAE/gdw, obtained from real experiments, demonstrated the validation of the extraction models. A comparison between the model results and experimental data gave high correlation coefficients ( R 2 ANN  = 0.999, R 2 RSM  = 0.981), adjusted coefficients ( R adjANN  = 0.999, R adjRSM  = 0.967) and low root mean square error s (RMSE ANN  = 0.37 and RMSE RSM  = 4.65) and showed that the two models were able to predict a total phenolic compounds (TPC) by green extraction ultrasound process. The results of ANN were found to be more consistent than DOE since better statistical parameters were obtained.