In order to improve the science and rationality of physical education, reduce various sports injury situations and improve students’ physical fitness. The prediction model method proposed in this paper organizes college and university students to participate in functional movement screening and selective functional movement assessment experiments to grasp the actual situation of students in physical exercise and to propose corresponding countermeasures. The prediction model takes advantage of the correlation nature of the time series, converts the vector autoregressive model into a linear regression model for research, then considers the graph regularization penalty function, combines this method with the bridge penalty, adds the correlation sign information between the variables, and uses the coordinate descent method for estimation, and finally proposes the vector autoregressive correlation prediction model method based on the bridge and the graph regularization. After the intervention-corrected training, the FMS scores of the college students were all improved, and the total score increased from the previous 12.39 to 17.51. Changes in students’ strength qualities before and after the experiment Except for pull-ups, there were significant interactions between students’ standing long jump scores, vertical long jump scores, grip strength, 1-minute push-ups, and 1-minute sit-ups in terms of time and group. This study led to an improvement in students’ physical functioning and reduced the risk of injury during physical activity.