id |
ddss9496 |
authors |
Veenendaal, Martin H. |
year |
1994 |
title |
Optimalization of Visualization: Graphical Diagonalization andClustering of Combinatorial Data |
source |
Second Design and Decision Support Systems in Architecture & Urban Planning (Vaals, the Netherlands), August 15-19, 1994 |
summary |
The analysis of combinatorial data is common to many disciplines as diverse as ethology, mathematics, computer science, psychology, demography, and architecture. Combinatorial data concern single relations that exist between the element pairs within one single pool of elements. TRI is a computer program that enables users to manually order combinatorial OTdangular) data matrices. "Ordering" in this context means placing high cell entries, coded as large dots, close together in clusters and close to the matrix's diagonal. Ordering, however, constitutes a very complex task. In order to support the ordering process, a straightforward measure has been developed which weighs the "amount" of clustering and diagonalization. The measure's value can be projected onto the monitor and perhaps serve as a "success indicator". A first experiment assessing the usefulness of the measure revealed that it does not consistently reflect subjective judgements of perceptual "order'. People may discern salient (although task irrelevant) patterns and regularities in dot configurations, for which the measure's cold calculus is insensitive [1]. In ongoing "human factors" experiments, the capability of experimental subjects to see through such "would be" order will be tested. One group will be amply instructed as to what the measure measures and how, and a second group will receive extensive visual instruction, using example matrices. The results of these and other experiments will help us decide whether or not to implement the measure in 1'RI, and how we can otherwise improve TRI as a powerful design and decision support tool. |
series |
DDSS |
email |
|
references |
Content-type: text/plain
|
last changed |
2003/08/07 16:36 |
|