This study used Xinjiang native "medicinal and food dual-use" resource mulberries as raw material, and optimized the extraction process of mulberries anthocyanins by enzyme-ultrasound-assistance through the establishment of a response surface model (RSM) and deep neural network model (DNN). A single-factor-Box-Behnken experiment was conducted to investigate the effects of pectinase dosage, enzymatic hydrolysis time, ultrasonic temperature, ultrasonic time, solvent concentration, and solid-liquid ratio on the extraction rates of total anthocyanins (TAC), and cyanidin-3-O-glucoside (C3G), cyanidin-3-O-rutinoside (C3R) two anthocyanin compounds, and the comprehensive evaluation index was used as a reference to obtain the optimal extraction conditions. The results show that both the RSM and DNN models could predict accurately, but by comparing the coefficient of determination (R