摘要: 综合考虑聚类、分类的特点,从聚类结果出发,学习并利用初始聚类结构信息形成训练集,结合迭代分类思想重新划分原数据集,提出一种基于迭代分类的聚类结果改进方法。实验结果表明该方法具有更高准确率,为获得良好的聚类效果提供了新思路。
关键词:
聚类,
聚类结果,
迭代分类,
k近邻分类
Abstract: This paper considers the features of clustering and classification synthetically. According to the clustering results, it constructs training sets by studying and utilizing structure information of initial clusters. The data sets are re-plotted according to iterative classification theory and an improved method based on iterative classification is proposed. Experimental results show that the method improves the accuracy of clustering results. It is a new way to obtain satisfactory clustering results.
Key words:
clustering,
clustering result,
iterative-classification,
k nearest neighbor classification
中图分类号:
王小华, 楼佳. 基于迭代分类的聚类结果改进方法[J]. 计算机工程, 2010, 36(13): 27-29.
WANG Xiao-Hua, LOU Jia. Clustering Result Improvement Method Based on Iterative-classification[J]. Computer Engineering, 2010, 36(13): 27-29.