This article studied the process of subcritical carbon dioxide isobaric extraction of chive seed oil. Box Behnken design was used for experimental design and optimization, and the effects of main parameters such as extraction pressure (11-23 MPa), temperature (50-70 ℃), and extraction time (60-150 minutes) on the experimental results were analyzed, the response surface method (RSM) and artificial neural network (ANN) were applied for modeling and predicting of the extraction yield. the performance of RSM and ANN models were analyzed and compared by statistical parameters such as coefficient of determination (R 2), root mean square error (RMSE), mean absolute error (MAE), and chi-square (χ 2). the RSM model is more accurate than the ANN model. Subsequently, optimization was carried by two different approaches viz. RSM and ANN-GA, by comparing the RSM and ANN models and the results of RSM and ANN-GA optimization, the RSM model is closer to perfection than the predictions of ANN. the optimal extraction conditions obtained by the RSM model were: extraction pressure of 15.63 MPa, separation temperature of 57.3°C, extraction time of 121.2 min,and predicted value 15.89%, mean value of three sets of parallel experiments 15.79%, the fatty acid composition analysis of leek seed oil optimized by RSM showed that linoleic acid (60.871%), oleic acid (19.185%), palmitic acid (11.517%) and stearic acid (3.174%) were the main components, and the content of four fatty acids was more than 94%.