The two models of microwave assisted aqueous two‐phase extraction (MATPE) of polysaccharides from Malus hupehensis are established by response surface methodology (RSM) and back propagation neural network (BP‐NN), compared their prediction ability, and comparatively optimized extraction process by using RSM and BP‐NN coupled genetic algorithm (GA), and the antioxidant activity of M. hupehensis polysaccharides (MHPs) is evaluated by the scavenging ability of free radicals. The results show that the predictive ability of BP‐NN is significantly better than that of RSM. The optimal extraction parameters to achieve the highest MHPs yield (4.71 ± 0.04)% by MATPE are obtained through BP‐NN coupled GA method under the microwave power of 337 W, mass fraction of ammonium sulfate of 23%, ethanol concentration of 29%, and liquid‐to‐solid ratio of 22 mL g−1. The IC50 values of 2,2‐diphenyl‐1‐picrylhydrazyl (DPPH) and hydroxyl radicals (OH) scavenging capacities for MHPs achieve 5.13 and 3.29 mg mL−1, respectively. The findings provide important references for the development and utilization of M. hupehensis.