Interactive Pattern Search in Time Series (28 Oct 2004)
Our early feedback highlighted the benefits of good browsing capabilities. Exploring the data visually is extremely important, and providing access to the numerical values needs to be supported with coordinated views of the graphs, tables and lists. Multiple presentations of the data are useful (for example overlapping versus sequential views of the items). User suggestions included providing multiple methods to specify the pattern and the options of the search algorithm. For example, one alternative to specifying tolerance would be to allow users to interactively modify the selected pattern or to draw boundaries of tolerance on the pattern by interacting with the line using direct manipulation. When the users know the pattern they are looking for and it is not easy to find an existing similar one, pure sketching may help15. The selection of patterns in existing data would also be complemented by a pure sketching option. This could easily be done with the current software. Alternatives can be offered as well for the selection of search parameters. Our early user feedback indicates that some users are confused by the tolerance slider label, because the tolerance measure has no real meaning to them. We initially chose to display a number to allow users to return to a previous value that had been found useful. Hopefully, the dynamic query behavior will help users quickly understand how to use the tolerance slider. An option might be to avoid displaying a numerical value and only provide + and – controls. Another natural suggestion is to extend the scope-setting Step 1 by allowing users to limit the time range where the search should be performed. Currently, nothing requires users to start with Step 1, and they can start with Step 2 and 3. Nevertheless it might be good to encourage users to reduce the scope of the search with the filter boxes when the size of the data doesn’t allow the use of dynamic queries.
Article URL: ftp://ftp.cs.umd.edu/pub/hcil/Reports-Abstracts-Bibliography/2004-25html/2004-25.pdf
Read 37 more articles from University Of Maryland sorted by
date,
popularity, or
title.
Next Article: KM for consistency & innovation
|