At present, there are still many limitations in the management of academic performance in colleges and universities. Most colleges and universities can master the learning situation of students only after the results are published. Students’ physical education is an important way to improve the physical quality of college students, and the prediction of college students’ physical education performance can help college sports management departments to reasonably set up relevant courses, develop the most scientific training mechanism. The more accurate the prediction of physical education performance, directly affects the formulation of training and combat preparation goals, and also affects the discovery of performance development laws and sports development characteristics. This paper constructs a prediction model of college students’ physical education performance based on BPNN (BP neural network) algorithm, and improves the simulation experiments to verify that the fitting degrees of the two methods are relatively close, but GA (Genetic algorithm) algorithm cannot predict the randomness in sprint performance, so the predicted value is higher than the actual value, and the prediction accuracy is inaccurate; The result of 200 m dash predicted by this algorithm is slightly better than that predicted by GA algorithm. The error between the 200 m dash result predicted by the algorithm in this paper and the actual value is lower, which shows that the algorithm in this paper has the best effect in predicting the students’ sports performance.