Creating Maps from Tables
|
Typically, users of VisuMap will begin with a tabular dataset. As an example, the right side shows the
key financial ratios of 104 US industry sections in Nov. 2003. Each
industry is characterized by eight numerical values (market capitalization,
price/earning ratio, etc.). VisuMap can help us, for instance, quickly
answer these questions:
- Which industry sections are similar to each other?
- Is an investment portfolio sufficiently diversified?
- How can the furture return of an investment portfolio be maximized?
|
US Industry Ratios
|
VisuMap implements
a collection of dimensionality-reduction methods to represent high-dimensional
datasets as low-dimensional maps.
The right side shows a map of the 104 US industry sections, as generated
by VisuMap's relational perspective map algorithm. Each spot in the
map represents an industry section. The area of each spot reflects that section's price/earning
ratio in that large spots represent industry sections with large price/earning
ratio. A major property of this map is that two industries with similar
financial ratios will be mapped to closely located positions. Thus, an
investment portfolio is well diversified if its corresponding industry
sections are widely distributed across the whole map.
|
RPM map of US Industry Ratios
|
VisuMap is designed as
a general purpose visualization tool for high dimensional datasets. Apart
from mapping and clustering algorithms, it offers a broad palette of
analysis services to explore datasets from different perspectives. The following
snapshots depict some of those services (click on these maps to see larger
images):
3D PCA |
MDS (Sammon)
|
Self-Organizing Map |
|
|
|
Value Diagram
|
Shepard Diagram
|
|
|
|
|
Spectrum with Value Diagram
|
3D Mountain View
|
|
|
More sample datasets can be found here.
|
