所属单位：计算机科学与技术学院/人工智能学院/软件学院
发表刊物：Commun. Comput. Info. Sci.
摘要：To address the issue that the K-means algorithm chooses and determines the initial cluster center in a random way, which would fall into the local optimal clustering result, a way towards choosing the initial clustering center using information entropy is proposed. This proposed method divides the dataset evenly into data blocks with more than K, and then uses the entropy method to obtain the value of target function of each data block, as well as selects the centroid corresponding to the data block with the smallest value function of the first k target as the initial cluster center. By using entropy method to ensure the efficiency of the initial clustering center selection, an anomaly detection method is proposed. The result of the experiment show that this method performs better than the traditional K-means algorithm both in clustering effect and anomaly detection ability. © Springer Nature Singapore Pte Ltd. 2019.
ISSN号：1865-0929
是否译文：否
发表时间：2019-01-01
合写作者：Fang, Xi,方贤德,Zhao, Lu,陆森召,Tang, Anqiong
通讯作者：谭文安,Fang, Xi,谭文安
发表时间：2019-01-01