摘要
判断聚类结果中是否存在误分类的簇,即簇中包含的样本不属于同一类。若存在,则在已有聚类结果上使用加权方案,处罚误分类的簇,输出新的聚类结果。若不存在,则输出已有聚类结果。限制簇集中存在误分类的簇,消除初始聚类中心对K-均值算法的影响,提高聚类准确率。实验结果表明,该算法与K-均值算法、优化初始聚类中心的K-均值算法相比,在坏的初始化条件下,表现出更好的鲁棒性;在含有噪音的数据集中,表现出更好的抗噪性能;聚类效果更好。
Judge the misclassification clustering existence in the K-means algorithm clustering or not,that is to mean the samples contained in the clustering do not belong to the same class. If existence,then utilizes a weighting scheme on existing clustering results,penalty for misclassification clustering to get the new clustering results. If not,to output the existing clustering results. To limit the misclassification clustering in the set and eliminate the influence of the initial clustering center of K-means algorithm would assure the clustering accuracy be advanced. The experimental results show that the proposed method is more robust in the unsatisfactory initialization conditions,have more noise immunity in a noisy data set,and clustering more accuracy to compare with the K-means algorithm,the K-means algorithm of optimized the initial clustering center.
出处
《长春理工大学学报（自然科学版）》
2015年第6期103-107,共5页
Journal of Changchun University of Science and Technology(Natural Science Edition)
关键词
聚类算法
K-均值算法
初始聚类中心
聚簇
clustering algorithm
K-means algorithm
initial clustering center
cluster
