Efficient extraction of flavonoid compounds in sweet tea (Lithocarpus litseifolius (Hance) Chun) via ultrasonic-natural deep eutectic solvent composite approach based on machine learning.

In this study, different types of natural deep eutectic solvents (NADES) were used to extract flavonoids from sweet tea. A total of 7 flavonoids, primarily dihydrochalcones, were successfully extracted. Among the NADESs tested, a ternary deep eutectic solvent Bet-PG-Ur (betaine: propanediol: urea = 1: 2: 1, molar ratio) demonstrated the highest extraction efficiency. The chemical and morphological changes of both the NADES and extracted compounds were analyzed using Fourier transform infrared spectroscopy (FT-IR) and scanning electron microscopy (SEM). To optimize the extraction process and evaluate the total flavonoid yield, response surface methodology (RSM), artificial neural network (ANN), and random forest (RF) techniques were employed. The NADES demonstrated consistent extraction performance even after three cycles of recovery using the SPE method. The results revealed that all three models successfully predicted the TFC in the green extraction ultrasonic process. The ANN model exhibited superior result consistency compared to the others, which can be attributed to the enhanced statistical parameters obtained. The highest total flavonoid content (TFC) of 167.561 mg/g was predicted by ANN and achieved under the following conditions: liquid/solid ratio of 19, water content of 50 %, temperature of 45 °C, and ultrasonic power of 400 W. [Display omitted] • A ternary DES-ultrasonic assisted was used for extraction of 7 flavonoids from sweet tea. • For the first time, machine learning was used to forecast the parameters for the TFC in sweet tea. • In terms of process modeling and prediction, ANN is better than RF and RSM. • The extraction efficiency of DES can reach 80% after 3 cycles of recovery by SPE. • Ultrasound-assisted extraction of 7 flavonoids from sweet tea using NADESs. [ABSTRACT FROM AUTHOR]