Objective To optimize the subcritical water extraction process of Pollen pini polysaccharides based on genetic algorithm-neural network (GA-NN) algorithm and further explore its anti-fatigue activity. Methods Using broken shell Pollen pini as raw material, a response surface was designed through Box-Behnken test on a single factor basis, and a network neural model was constructed and optimized using GA-NN algorithm. Under the optimal process conditions, the anti-fatigue effect of polysaccharides was evaluated through mouse weight-bearing swimming test. Results The relative error and coefficient of determination (R²) of the constructed neural network model were 0.03267 and 0.98476, respectively. The genetic algorithm was iterated 60 times for subcritical water extraction of polysaccharides, and the optimal parameters were obtained as follows: Temperature 148 ℃, time 28 min, liquid to material ratio 40:1 (mL/g), pressure 5 MPa, and polysaccharide yield of 23.7893 mg/g. After verification, there was no significant difference between the actual value and the predicted value, indicating good accuracy of the model. The study on anti fatigue activity showed that compared with the blank group, the high-dose group of mice had a certain degree of increase in body weight, while there was no significant difference in other groups (P>0.05). Compared with the blank group, the swimming time of mice in the polysaccharide group was prolonged, the levels of blood lactate and urea nitrogen were reduced, and the reserves of muscle glycogen and liver glycogen were significantly increased (P<0.05). Conclusion GA-NN can effectively optimize the subcritical water extraction process of Pollen pini polysaccharide, and the polysaccharide has a certain anti fatigue effect.