Unsupervised Ensemble Learning

Clustering is used in identifying groups of samples with similar properties, and it is one of the most common preliminary exploratory analysis for revealing hidden patterns, in particular for datasets where label information is unknown. Even though clustering techniques have been well used to analyze a variety of datasets in different domains for years, the limitation of them is that each clustering method works better only in certain conditions. This made the selection of the most suitable algorithm for particular dataset much more important. Restrained implementation of clustering methods has forced clustering practitioners to develop more robust methods, which is reasonably practicable in any condition. The unsupervised ensemble learning, or consensus clustering, is developed to serve this purpose. It consists of finding the optimal combination strategy of individual partitions that is robust in comparison to the selection of an algorithmic clustering pool. The goal of this combination process is to improve the average quality of individual clustering methods. Due to increasing development of new methods, their promising results and the great number of applications, it is considered to make a crucial and a brief review about it. Through this chapter, first the main concepts of clustering methods are briefly introduced and then the basics of ensemble learning is given. Finally, the chapter is concluded with a comprehensive summary of novel developments in the area