A Semi-supervised Clustering Algorithm Based on Must-Link Set

摘要

Clustering analysis is traditionally considered as an unsu-pervised learning process. In most cases, people usually have some prior or background knowledge before they perform the clustering. How to use the prior or background knowledge to imporve the cluster quality and promote the efficiency of clustering data has become a hot research topic in recent years. The Must-Link and Cannot-Link constraints between instances are common prior knowledge in many real applications. This paper presents the concept of Must-Link Set and designs a new semi-supervised clustering algorithm MLC-KMeans using Musk-Link Set as assistant centroid. The preliminary experiment on several UCI datasets confirms the effectiveness and efficiency of the algorithm.