Visualizing Uncertainty in Predictive Models - Scientific Visualization: Uncertainty, Multifield, Biomedical, and Scalable Visualization

Image Processing Reference

In-Depth Information

centerline of the visualization, there is an individual whose true class is tech-support

(red), but who is predicted to be either in the sales (brown) or exec-managerial (dark

green) classes).

It is not immediately obvious whether the speckle or pie chart glyph shapes are

“better” for human understanding in the general case. It does seem to be the case that

when several different classes are predicted, this pattern may be slightly easier to see

using the speckle glyphs. We are currently designing a user study to investigate the

validity of this hypothesis.

6.4 Future Work and Conclusions

Our research is in its early stages, but our preliminary results show promise for the

use of integrated machine learning and visualization techniques to improve the repre-

sentation and human understanding of uncertainty in classification models. Specific

future work includes designing and carrying out a user study of the alternative visu-

alization techniques, performing an empirical investigation of alternative dimension

reduction techniques, and investigating machine learning methods for improving the

confidence estimates associated with the predictive models.

Acknowledgments This preliminary work was supported by NSF EAGER #1050168 and by NSF

REU Supplement #1129683. Thanks to David Mann for his contribution to the project.

References

1. Assa, J., Cohen-Or, D., Milo, T.: Displaying data in multidimensional relevance space with 2D

visualization maps. In: IEEE Visualization '97, pp. 127-134. IEEE Computer Society Press,

Washington, D.C. (1997)

2. Cedilnik, A., Rheingans, P.: Procedural annotation of uncertain information. In: Proceedings

of IEEE Visualization '00, pp. 77-84 (2000)

3. Cox, T., Cox, M.: Multidimensional Scaling. Monographs on Statistics andApplied Probability.

Chapman & Hall, London (1994)

4. Davis, T.J., Keller, C.P.: Modelling and visualizing multiple spatial uncertainties. Comput.

Geosci. 23 (4), 397-408 (1997)

5. Hagh-Shenas, H., Kim, S., Interrante, V., Healey, C.: Weaving versus blending: a quantita-

tive assessment of information carrying capacities of two alternative methods for conveying

multivariate data with color. IEEE Trans. Visual. Comput. Graph. 13 (6), 1270-1277 (2007)

6. Hall, M., Frank, E., Holmes, G., Pfahringer, B., Reutemann, P., Witten, I.H.: The WEKA data

mining software: an update. SIGKDD Explor. 11 (1), 10-18 (2009)

7. Johnson, C., Sanderson, A.: A next step: visualizing errors and uncertainty. IEEE Comput.

Graph. Appl. 23 (5), 6-10 (2003)

8. Joshi, A.: Art-inspired techniques for visualizing time-varying data. Ph.D. thesis, University

of Maryland, Baltimore County (2007)

9. Kohavi, R., Becker, B.: Adult data set (1994). Available online from UCI Machine Learning

10. Kohonen, T.: Self-organizing Maps. Springer, Berlin (1995)

Search WWH ::

Custom Search

Home