iVisClustering: An Interactive Visual Document Clustering via Topic Modeling

摘要: Clustering plays an important role in many large-scale data analyses providing users with overall understanding of their data. Nonetheless, clustering is not easy task due to noisy features and outliers existing the data, thus results obtained from automatic algorithms often do make clear sense. To remedy this problem, should be complemented interactive visualization strategies. This paper proposes visual analytics system for document clustering, called iVisClustering, based on a widely-used topic modeling method, latent Dirichlet allocation (LDA). iVisClustering provides summary each cluster terms its most representative keywords visualizes soft parallel coordinates. The main view 2D plot that similarities relation among items graph-based representation. several other views, which contain useful interaction methods. With help these modules, we can interactively refine various ways. Keywords adjusted so they characterize better. In addition, our filter out re-cluster accordingly. Cluster hierarchy constructed using tree structure purpose, supports cluster-level interactions such as sub-clustering, removing unimportant clusters, merging clusters have similar meanings, moving certain any node structure. Furthermore, document-level mis-clustered documents another useless documents. Finally, present how performed via by real-world sets. © 2012 Wiley Periodicals, Inc.