Research on domain-specific features clustering based spectral clustering

Abstract

Domain-Specific features clustering aims to cluster the features from related domains into K clusters. Although traditional clustering algorithms can be used to domain-specific features clustering, the performance may not good as the features have little inter-connection in related domains. In this paper, we develop a solution that uses the domain-independent feature as a bridge to connect the domain-specific features. And we use spectral clustering to cluster the domain-specific features into K clusters. We present theoretical analysis to show that our algorithm is able to produce high quality clusters. The experimental results show that our algorithm improves the clustering performance over the traditional algorithms. © 2012 Springer-Verlag.