半监督聚类利用已标记的数据样本对聚类过程进行指导,提高了无监督学习的准确率,但是现有的半监督聚类算法都是针对完备标签数据提出的,这样的要求不切合实际的应用。提出一种新的半监督聚类算法,首先通过自适应的方法预估聚类数,然后利用优化目标函数方法来寻求最佳聚类中心。该方法可以对不完备标签数据进行聚类划分,而且降低计算复杂度,实验验证其聚类结果和计算复杂度都有明显的改善。
emi-supervised clustering uses labeled data samples to guide the clustering process and improve the accuracy of unsupervised learning.But the existing semi-supervised clustering algorithms are proposed for complete label data,such a request does not meet the actual application.This paper proposes a novel semi-supervised clustering algorithm.It firstly applies the adaptive method to estimate the number of clusters,then uses the method by optimizing the objective function to find the best cluster centers.