The objective of this research was to investigate the efficacy of 10 NDES, including 3 NDES with new formula, to extract procyanidins and anthocyanins from cranberry pomace with the assistance of ultrasounds. The highest amount of procyanidins (32.5 mg/g) was extracted by a tailor designed NDES 2 consisting of choline chloride: betaine hydrochloride: levulinic acid (1:1:2) and 32 mL water/100 mL NDES. This yield was 3.26-fold of that by 75% ethanol. NDES 8 consisting of glucose: lactic acid (1:5) and 20 mL/100 mL water had the highest extraction yield of anthocyanins at 1.58 mg/g, which was 1.79-fold of the yield by 75% ethanol. The response surface methodology model for extraction of procyanidins by NDES 2 under different conditions had R² = 0.977, which was comparable to artificial neural networking (R² = 0.973). The ANN models for extraction of anthocyanins using NDES 8 under various conditions performed better than RSM model (R² = 0.95 for ANN versus 0.88 for RSM). Highest extraction yields predicted by ANN and RSM were close to the experimental yields, suggesting artificial neural networking was an alternative or better approach than RSM for predictive modeling.

这项研究的目的是研究10种NDES（包括3种新配方的NDES）在超声波的辅助下从蔓越莓果渣中提取原花青素和花色苷的功效。通过量身定制的NDES 2提取最高量的原花青素（32.5 mg / g），该NDES 2由氯化胆碱：甜菜碱盐酸盐：乙酰丙酸（1：1：2）和32 mL水/ 100 mL NDES组成。该产率是75％乙醇的产率的3.26倍。由葡萄糖：乳酸（1：5）和20 mL / 100 mL水组成的NDES 8的花青素提取率最高，为1.58 mg / g，是75％乙醇的提取率的1.79倍。在不同条件下用NDES 2提取原花青素的响应面方法学模型的R ²  = 0.977，可与人工神经网络（R²  = 0.973）。在各种条件下使用NDES 8提取花色苷的ANN模型的性能优于RSM模型（ ANN的R ² = 0.95，RSM的0.88）。ANN和RSM预测的最高提取产量接近于实验产量，这表明在预测建模方面，人工神经网络是一种比RSM更好的替代方法。