There are many factors affect athletics competition, so it needs better prediction model to make prediction on athlete each event performance, the paper just based on such thought, it introduces BP neural network model, RBF neural network model and Elman neural network model, after comparing, it gets that BP neural network of the three has higher accuracy, in addition, by combining the three kinds of neural network models and then introduces it into specific examples, it gets that model after combining accuracy is higher than any one kind of single prediction model, so it proves such model’s superiority, it will play important driving roles in neural network’s sports performance prediction development.