Research on Improving Higher Education Exam Quality Based on Weighted k-Medoids Clustering

Exam is the most effective method to evaluate the quality of higher education, and improving higher education exam quality is of paramount importance. Traditional methods of analyzing and improving higher education exam quality, such as mean and variance based on mathematical statistics, are only suitable for sample datasets that are small and static. On the other hand, robust clustering methods, such as PAM and k-Medoids, do not consider the importance of each attribute which lead to different impacts on clustering result. Based on the aforementioned issues, this paper researches on improving higher education exam quality based on weighted k-Medoids clustering. Specifically, the calculating method of attribute weights is introduced based on the granularity rough entropy. Secondly, a novel weighted k-Medoids clustering method is proposed, which integrates the attribute weights into the classic k-Medoids clustering method. Finally, the performance on UCI datasets shows the proposed method significantly improves the clustering accuracy compared to PAM and fast k-Medoids. Meanwhile, the experimental results on proprietary artificial teaching datasets indicate that the novel method identifies and corrects redundant and less significant exam questions, effectively improving higher education exam quality.