Risk early warning analysis of Asian Games contestants based on improved BP neural network

Abstract Sports is a complex sports system, which fundamentally determines the objectivity and randomness of sports risks, and risks can often be predicted by certain methods. The occurrence of any risk event has its inevitability and chance, usually the result of the accumulation of risk factors to a certain stage. The occurrence of risks often needs to meet certain conditions and taking necessary means to control the existing risk factors is an effective means to reduce risks. In this study, a BP neural network model for risk early warning of Asian Games players was established, and the constructed Asian Games players' participation risk early warning model was effective and had high prediction accuracy. The network training accuracy of the early warning model is 0.003328, indicating that the model can explain about 99% of the data samples. Between the actual value and the predicted value of the model prediction results, except for individual absolute errors reaching 0.09, the remaining errors are all less than 0.05. In addition, the predicted risk of participating in the Asian Games is mainly concentrated in the two levels of medium police and light police , with light police accounting for 35%.