The scale and complexity of sports events put forward higher requirements for the safety of athletes and the guarantee of the fairness of the competition. In this context, intelligent risk identification has become a key technology to improve the security of sports systems. Association analysis algorithms, especially Frequent Pattern Growth (FP-Growth) algorithms, are widely used in risk identification because of their effectiveness in mining interesting association rules in data. This paper firstly collects and preprocesses sports data sets, and uses FP-Growth algorithm to mine risk patterns in the data sets. In this paper, the performance of the algorithm is evaluated comprehensively by calculating risk recognition rate, risk response time, false positive rate and user satisfaction. Experimental results show that the risk recognition rate of FP-Growth algorithm reaches 89% at the highest, the risk response time only takes 133ms at the highest, the false positive rate ranges from 1.01% to 1.19%, and the user satisfaction is relatively high. Future research will continue to explore further optimization and fusion of algorithms and how to improve the explainability of algorithms to enhance user trust and satisfaction with the system.