Data mining has been applied to a large extent in various information application fields this year. In this paper, it is applied to the students' physical exercise for partial optimization and compared with some traditional algorithms for improvement. Different analysis results are obtained under different potential information analysis backgrounds, which improves the accuracy and results of clustering analysis. This study used the physical activity rating scale and the subjective well-being scale to assess the status, subjective well-being level, and interaction mechanism of middle school students in Sichuan Province participating in physical exercise. The physical activity of students is insufficient, and the physical activity of girls is seriously insufficient; the overall level is above average, and the average score of boys in the two dimensions of life satisfaction and emotional balance is higher than that of girls; the relationship between each dimension of physical exercise and life satisfaction and emotional balance is shown.