Interactive Visual Overviews of Large, Multi-dimensional Datasets

 

Personnel

Dr. Dianne Cook , Associate Professor of Statistics, Principal Investigator

Dr. Vasant Honavar, Professor of Computer Science and of Bioinformatics and Computational Biology, Co-Principal Investigator

Dr. Les Miller , Professor of Computer Science, Co-Principal Investigator.

 

Summary

This research was aimed at development and integration of dynamic graphics tools with machine learning algorithms for exploratory analysis, interactive exploration, and visualization of large, multi-dimensional data. The resulting tools enable (Caragea, Cook, and Honavar, 2001; Caragea, Cook, and Honavar, 2003; Cook, Caragea, and Honavar, 2004):

• Visualization of complex relationships discovered from large data sets using machine learning algorithms (e.g., support vector machines) Detection of detect structure in high-dimensional data, and local departures from an overall trends in the data,
• Selection of data-dependent preprocessing steps (e.g., scaling, normalization, feature extraction) to enhance the effectiveness of machine learning algorithms,
• Refinement of solutions (e.g., decision surfaces for classification) obtained with machine learning algorithms,
• Obtaining overviews the data space as well as the complex relationships discovered from the data and gain new insights into the working of popular machine learning algorithms, and the use the resulting insights to develop improved algorithms.

 

Funding

This work was funded in part by a National Science Foundation Grant ACI 9982341.

Representative Publications

1. Cook, D., Caragea, D., and Honavar, V. Visualization in Classification Problems. Proceedings in Computational Statistics (COMPSTAT 2004), Springer-Verlag. pp. 799-806, 2004.
2. Caragea, D., Silvescu, A., and Honavar, V. A Framework for Learning from Distributed Data Using Sufficient Statistics and its Application to Learning Decision Trees. International Journal of Hybrid Intelligent Systems. Vol. 1. No. 2. pp. 80-89, 2004.
3. Caragea, D., Cook, D., and Honavar, V. Toward Simple, Easy-to-Understand Classifiers. Proceedings of the IEEE International Conference on Data Mining, IEEE Press. pp. 497-500, 2003.
4. Caregea, D., Cook, D., & Honavar, V. Visualizing Ensemble of Hyperplane Classifiers. IEEE International Conference on Data Mining, Melbourne, Florida, Springer Verlag, 2003.
5. Caragea, D., Cook, D., and Honavar, V. Gaining Insights into Support Vector Machine Classifiers Using Projection-Based Tour Methods. ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco, CA, USA, ACM. pp. 251-256, 2001.
6. Caragea, D., Silvescu, A., and Honavar, V. Towards a Theoretical Framework for Analysis and Synthesis of Agents That Learn from Distributed Dynamic Data Sources. In: Emerging Neural Architectures Based on Neuroscience (Ed. Wermter, S., Austin, J. & Willshaw, D.), Springer-Verlag 2001.