05.06.2013 Views

PNNL-13501 - Pacific Northwest National Laboratory

PNNL-13501 - Pacific Northwest National Laboratory

PNNL-13501 - Pacific Northwest National Laboratory

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Figure 1. A visualization technique to explore sequential<br />

patterns<br />

Figure 2 presents a novel technique to visualize many-toone<br />

association rules. The rows of the matrix floor<br />

represent the topics, and the columns represent the item<br />

associations. The blue and red blocks of each column<br />

(rule) represent the antecedent and the consequent of the<br />

rule. The identities of the items are shown along the right<br />

side of the matrix. The confidence and support levels of<br />

the rules are given by the corresponding bar charts in<br />

different scales at the far end of the matrix. The system<br />

supports basic query commands through the use of pop-up<br />

menus to restrict key items to be included in the<br />

visualization.<br />

In Figure 3, each data record of a multivariate data set is<br />

represented as a series of line segments that together<br />

create what we refer to as a line plot. Each line segment<br />

reflects two measures of a characteristic of that data<br />

record. Line plots are placed next to each other in the<br />

visualization such that when no variation occurs in their<br />

line segments, they are parallel with one another. When<br />

used to represent a multivariate data set, the line plots<br />

visually represent the degree to which the derived<br />

measures for each parameter vary.<br />

Figure 2. A visualization technique to explore association<br />

rules<br />

Figure 3. A visualization technique to explore anomalous<br />

exceptions of a large multivariate dataset<br />

Summary and Conclusions<br />

This project presents data mining and visualization<br />

techniques for discovery of patterns and exceptions from<br />

large corpora. The strengths of the two approaches (data<br />

mining and visualization) can compensate for each other’s<br />

weakness. We subsequently introduced a powerful visual<br />

data-mining environment that contains multiple datamining<br />

engines to discover different patterns and<br />

exceptions, and multiple visualization front-ends to show<br />

the distribution and locality of the mining results. Our<br />

experimental results show that we can quickly learn more<br />

in such an integrated visual data-mining environment.<br />

Publications and Presentations<br />

Wong PC, W Cowley, H Foote, E Jurrus, and J Thomas.<br />

2000. “Visualizing sequential patterns for text mining.”<br />

Presented and In Proceedings IEEE Information<br />

Visualization 2000, October 2000, IEEE Computer<br />

Society Press, Los Alamitos, California.<br />

Rose S and PC Wong. 2000. “DriftWeed - A visual<br />

metaphor for interactive analysis of multivariate data.”<br />

Proceedings IS&T/SPIE Conference on Visual Data<br />

Exploration and Analysis, January 24 - 26, 2000, San<br />

Jose, California.<br />

Wong PC, P Whitney, and J Thomas. 1999. “Visualizing<br />

association rules for text mining.” Presented and In<br />

Proceedings IEEE Information Visualization 99, October<br />

1999, IEEE Computer Society Press, Los Alamitos,<br />

California.<br />

Wong PC. 1999. “Visual data mining.” IEEE Computer<br />

Graphics and Applications, 19(5).<br />

Computer Science and Information Technology 163

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!