Latent semantic indexing: a probabilistic analysis
PODS '98 Proceedings of the seventeenth ACM SIGACT-SIGMOD-SIGART symposium on Principles of database systems
Concept decompositions for large sparse text data using clustering
Machine Learning
Database-friendly random projections
PODS '01 Proceedings of the twentieth ACM SIGMOD-SIGACT-SIGART symposium on Principles of database systems
Random projection in dimensionality reduction: applications to image and text data
Proceedings of the seventh ACM SIGKDD international conference on Knowledge discovery and data mining
Modern Information Retrieval
Self-Organizing Maps
In-depth behavior understanding and use: The behavior informatics approach
Information Sciences: an International Journal
Experiments with random projection
UAI'00 Proceedings of the Sixteenth conference on Uncertainty in artificial intelligence
Visualization of cluster changes by comparing self-organizing maps
PAKDD'05 Proceedings of the 9th Pacific-Asia conference on Advances in Knowledge Discovery and Data Mining
| Hi-index | 0.00 |
In this paper, a novel method for visualizing cluster structures and their changes over time is proposed. Clustering is achieved by two-step application of self-organizing maps (SOMs). By two-step application of SOMs, each cluster is assigned an angle and a color. Similar clusters are assigned similar ones. By using colors and angles, cluster structures are visualized in several fashions. In those visualizations, it is easy to identify similar clusters and to see degrees of cluster separations. Thus, we can visually decide whether some clusters should be grouped or separated. Colors and angles are also used to make clusters in multiple datasets from different time periods comparable. Even if they belong to different periods, similar clusters are assigned similar colors and angles, thus it is easy to recognize that which cluster has grown or which one has diminished in time. As an example, the proposed method is applied to a collection of Japanese news articles. Experimental results show that the proposed method can clearly visualize cluster structures and their changes over time, even when multiple datasets from different time periods are concerned.