摘要：Semi-supervised clustering which uses the limited labeled data to aid unsupervised clustering, has become a hot topic in recent years. But the limited labeled data may be imbalanced and can not cover all clusters in some cases and most of the existing semi-supervised clustering algorithms can not deal with imbalanced dataset well and have no the ability of detecting new clusters. In view of this, an adaptive semi-supervised clustering algorithm with label propagation is proposed. Two most of interesting characteristics of the proposed algorithm are that (1) It uses the limited labeled data to expand labeled dataset based on an adaptive threshold by labeling their k-nearest neighbors, (2) It detects whether there exist new clusters in the unlabeled dataset according to a proposed measure criterion. Three standard datasets are used to demonstrate the performance of the proposed algorithm and the experimental results confirm that the accuracy of the proposed clustering algorithm is much higher than that of three compared algorithm and in addition the proposed algorithm has the ability of detecting new clusters.
