Graphics Reference
In-Depth Information
to be very flexible to accommodate all of these different criteria. Ideally, one would
want to literally “dive" into the data and explore it interactively.
Information visualization tools apply several common strategies that enable user
control over data displays(see Shneiderman and Plaisant (
),Card etal.(
)or
Plaisant(
)).Aprimarystrategyistoallowtheusertomanipulateasetofwidgets,
such as dynamic query sliders that allow the user to select the ranges of the desired
variables, in aprocessoten called conditioning. hepowerof interaction is that users
can rapidly (
ms) and incrementally change the ranges and explore the effect of
these changes on the display.Forexample, userscan moveaslidertogradually elimi-
nateauctionswithlowstartingpricesandseeifthatremovestimeseriesplotsthatend
with low, middle, or high closing prices. A second strategy is to have multiple views
of the data, suchas scattergram, histogram, tabular, or parallel coordinate views. he
users can then select a single or multiple items in one view and see the results in
another view (“brushing"). For example, users can select the time series with sharp
increases near the finish in order to see if these had relatively few previous bids.
Selectivity and user control are essential, as they support exploration (to confirm
hypotheses) and discovery (to generate new hypotheses) (Chen,
). he large
number of possibly interesting features in high-dimensional data means that static
displays and a fixed set of data-mining algorithms may not be enough. Users can
quicklyspotunusualoutliers,bimodaldistributions, spikes,longorshorttails onone
side of a distribution, and surprising clustersor gaps.Users mayalso detect strong or
weak relationships, which can be positive or negative, linear, quadratic, sinusoidal,
exponential, etc.
he strongest tools are likely to be those that combine data-mining algorithms
with potent userinterfaces (Shneiderman,
).hesehave the potential toprovide
thorough coverage through a systematic process of exploration in which users can
decompose a complex problem domain into a series of simpler explorations with
ranking criteria, and they guide user attention to potentially interesting features and
relationships (Seo and Shneiderman,
).
Interactive Information Visualization
of Functional and Cross-sectional
Information via TimeSearcher
5.5
TimeSearcher is a time series visualization tool developed at the Human-Computer
Interaction Laboratory (HCIL) of the University of Maryland. TimeSearcher enables
users to see an overview of long time series (
points), to view multivariate
time series, to select data with rectangular time boxes, and to search for a selected
pattern. Its main strength comes from its interactivity, which allows users to explore
time series data in an active way.Unlike static graphs, an interactive approach can be
more powerful and can lead to a better understanding of the data.
