Abstract: K-means clustering method is one of the classic and widely-used clustering analysis methods because of its fast execution efficiency and simple algorithm.However, the traditional K-means clustering algorithm does not consider the importance of each index while the importance of each index is different and it should be treated differently. This paper improves the traditional K-means clustering method and proposes a weighted K-means clustering algorithm in order to verify the superiority of the improved algorithm, we take iris as the experimental data, and set up three experiments for comparative analysis. The experimental results show that the weighted K-means clustering algorithm converges faster, the classification result is closer to the original classification, and the effect is better. In order to further illustrate the practicability of weighted K-means clustering algorithm, this paper takes some indexes of 103 colleges and universities as experimental data, and uses the algorithm for clustering analysis.

Key words: K-means clustering, weighted K-means clustering, standard deviation coefficient method, university clustering