To investigate the optimal green extraction process and antioxidant activity of total flavonoids from Qinling Artemisia argyi，the optimization performances of response surface methodology (RSM) and BP neural network (BPNN) on ultrasound-assisted natural deep eutectic solvents (NADES) extraction parameters were comparatively analyzed based on single-factor experiments，using total flavonoid yield as the evaluation index.Furthermore，HPLC and in vitro free radical scavenging assays were employed to comprehensively evaluate the compositional differences and antioxidant efficacy of the extracts.The results indicated that betaine-urea was the most suitable NADES system.Compared with RSM，the BPNN model demonstrated superior non-linear fitting and optimization capabilities.Under the optimal conditions predicted by BPNN (ultrasonic temperature of 80 ℃，molar ratio of 1.01∶1，and water content of 38%)，the total flavonoid yield reached 177.43 mg/g.Compositional and bioactivity analyses further revealed that，due to solvent polarity selectivity，although the contents of jaceosidin and eupatilin in the NADES extract were lower than those in the traditional ethanol extract，the NADES extract exhibited stronger antioxidant activity，with IC50 values for DPPH and ABTS radical scavenging as low as 0.041,0.11 mg/mL，respectively.This study provides a theoretical foundation for the green and efficient extraction of Qinling Artemisia argyi resources and their development and application in natural antioxidant products.