SummaryChina has continuously strengthened its efforts in reforming the sports training and competition organization mode, issued a series of measures and made some innovations, but the current optimized policies and measures still have certain limitations. Therefore, this article first uses data mining technology to model students' sports training and competition mode, considering multi event competition, the scoring of the project itself, the uncertainty of the number of students in sports training and other factors. The results show that the average prediction accuracy of the time series data mining method applied in this article is 89.82%, which is 19.71% and 44.78% higher than that of ANN and traditional clustering methods, respectively. The average error rate of ARMA time series algorithm data mining is reduced by 21.95% and 18.89% compared with AR and Ma time series algorithm. In addition, compared with the convergence time of multidimensional sports data mining realized by AR algorithm and Ma algorithm, ARMA time series algorithm is shortened by 77.13% and 70.02%, respectively.