In recent years, with China’s economic and social development, higher education has developed rapidly, and college physical education courses have received more and more attention, so the research on the evaluation system of college physical education courses is particularly important. Based on the perspective of Hefei University of Technology, this paper integrates a K-means clustering algorithm to optimize the evaluation system of college sports courses. By analyzing the problems existing in the current evaluation system of physical education courses at Hefei University of Technology, a method of quantitative evaluation of “course quality” by calculating the evaluation value of course attainment and course attainment rate is studied. It also proposes to use the K-means clustering mining algorithm as the mining algorithm for the evaluation of university physical education courses, improve the algorithm, and then apply it to the evaluation analysis of university physical education courses. By analyzing the quality of physical education courses at Hefei University of Technology, the total percentage of failing and poorer students is 8%, which indicates that the quality of physical education courses at Hefei University of Technology is better. The optimized evaluation system of college physical education courses is practiced and taught, and the results show that the optimized evaluation system of college physical education courses has a positive effect relative to the traditional evaluation mode from the analysis of student’s interest and attitude towards course evaluation.