|
Abstract
|
The conventional evaluation framework for university physical education (PE) often overlooks the unique differences and varied needs of students, limiting its ability to fully capture the instructional quality. This research introduces an enhanced evaluation system for university PE that seeks to improve its scientific validity and overall effectiveness. The study enhances the traditional K-modes algorithm by integrating a co-occurrence rate distance measure, thereby improving clustering precision and minimizing noise sensitivity. An empirical study was conducted using student evaluation data from the School of Physical Education at a university, covering the academic years from 2014 to 2023. The findings reveal that the refined K-modes algorithm notably surpasses both the original K-modes algorithm and the frequency-weighted variant in terms of Sum of Squared Errors (SSE), accuracy, and recall rate. Specifically, the enhanced algorithm recorded an SSE of 598, which is significantly lower than the 1374 of the original K-modes algorithm and 841 of the frequency-weighted version. Additionally, the accuracy of the improved algorithm was 0.9784, significantly higher than the 0.8724 achieved by the traditional algorithm and the 0.9523 of the frequency-weighted algorithm. The recall rate of 0.9869 further highlights the algorithm’s superior ability to correctly identify classified data. Moreover, the study identified that among the four evaluation dimensions—teacher professionalism, teaching attitude, teaching skills, and extracurricular activities—the 2023 evaluation results were relatively consistent, with more than 75% of ratings in professionalism and teaching attitude falling within excellent or good categories. However, there was a marked divergence in teaching skills and extracurricular activities. These results offer both theoretical support and practical evidence for enhancing university PE quality and provide key insights for educational management decision-making.
|