Recently, nutrition researchers have been aware of the healing properties of apricot kernel and apple seeds as a source of fiber, minerals, antioxidants, unsaturated fatty acids, and its effect on coughing, constipation, asthma, bronchitis and other inflammation. And autoimmune diseases, cancer and etc have proven themselves. Although many antioxidants have been mentioned for them, many believe it and have proven their anti-tumor and anti-cancer properties through the presence of a substance called amygdalin or vitamin B-17. In this study,first, amygdalin extraction from apple seed and apricot kernel was performed by Soxhlet and methanol was used as solvent for this method. Then, with the aid of the Minitab17 software, experiments were designed for 4 operational parameters of temperature, pressure, carbon dioxide flow rate and dynamic time in 5 levels with CCR design and response surface methodology, and performed for supercritical carbon dioxide extreaction, taking into account the 30minute as static time and using methanol as a cosolvent. The range of operating parameters for each of the parameters is as follows: temperature (40-80 ° C ), pressure (10-34 MPa), carbon dioxide flow rate (1-2.2 ml / min) and dynamic time (40-120 ° C) Min). Then, the results were analyzed with the help of Minitab 17 . The model parameters include linear and square terms with significant effect and some interaction parameters. The recovery fraction of amygdalin in two matrices and different conditions were formulated as a second-order polynomial with more than 90% of R 2 and an adjusted of R 2 of more than 85%. Maximum recovery fraction of amygdalin for extraction from apricot kernel apple seed was 0.58 and 0.40 for, respectively. In RSM, optimum operational conditions for extraction of amygdalin from apricot kernel are as 80° C,25 Mpa, 1.7 ml/min and 120min, as well as extraction from apple seed are 80° C,20 Mpa,1.15 ml/min and 120min. In mathematical modeling based on the differential mass balance model, model parameters, which include effective diffusion, film mass transfer coefficient of fluid ,partitioning coefficient and axial dispersion were calculated using empirical equilibrium relatio and experimental data. Then, using the model based on these relationships and parameters and the genetic algorithm, the maximum recovery rate and optimal parameters were calculated. In bocken and intact cell model, the model parameters, which include volumetric mass transfer of solid phase and grinding efficiency, were calculated by three adjustable parameters determined by exoerimental data. At the end, the process was modeled using a multi-layered perceptron artificial neural network by assigning 3 neurons in the hidden layer for extraction data of apricot kernel and 8 neurons for modeling the apple seed extraction data this model could satisfactorily comply more than 98% with experimental data. Keywords : Amygdalin, Supercritical extraction, Response surface method, Mathematical modeling, Broken and intact cell model, Genetic Algorithm, MLP Artificial Neural Network