用聚类-分类模式解决聚类问题

Clustering Based on Clustering-Classification Model

分类和聚类都是常用的数据挖掘方法,分类的优点是准确率较高,但需要带有类别标注的训练集;聚类不需要训练集,但准确率较低。提出一种聚类-分类模式来解决聚类问题,首先通过聚类方法自动形成训练集,然后在训练集的基础上进行分类操作。实验数据表明,提出的聚类-分类模式能够有效提高聚类的准确率。

Classification and clustering are both commonly used data mining methods. The advantage of classification is that the accuracy is higher ,but the labeled training set is needed. The training set is not needed in clustering but the accuracy is lower. A clustering-classification model is proposed to solve the clustering problem. First ,the training set is formed automatically by clustering, and then the classification proceeds based on the training set. Experiments show that the cluster-classification model can improve the effect of clustering.