Information Technology Reference
In-Depth Information
occurrence in time. For example, we can consider a web access log as data
with temporal information, because the log records which pages were
accessed and at what time. Generally, an event occurs once or more, and
then has one or more time-stamps. Suppose that access to a web page is an
event. If a web page is accessed at 9 p.m., 10 p.m., and 11 p.m., the event
(that is, the access to the web page) has three time-stamps. Many different
types of such data exist, besides web access logs. Other examples include
product sales histories, management data such as Subversion (SVN), and
incident logs at a call centre.
The purpose of our research is to support the analysis of large amounts
of temporal data. In particular, we focus on the set of time-stamps
detailing when an event occurred. As described above, log data can
include many events that are associated with one or more time-stamps. By
analysing sets of large numbers of time-stamps, we can obtain valuable
knowledge about the occurrence of events. For instance, we can determine
trends in a phenomenon, periodicity, causality, and so on. Such knowledge
is useful in fields such as market research, security, and resource
management.
Data visualization can be streamlined to help the analysis of data. A
number of visualization techniques for temporal information have been
proposed [1]. Many current visualizations represent the time axis by a
straight line and plot the individual time-stamps on the line segment. To
represent periodicity, some visualizations use a circumferential or spiral
representation of the time axis instead of straight-line segments [2]. In
either approach, visualizing a set of time-stamps needs a certain area.
Therefore, the existing visualization techniques have some problems with
scalability. First, we cannot have an overview of all the events at once. For
example, if we want to look at data containing at least 1,000 events, a
normal display cannot fit all of these, and we cannot avoid scrolling
horizontally or vertically. Second, even if we could place all events on the
display, it is difficult to understand their characteristics. This kind of
representation does not support the analysis of events with similar
temporal characteristics or the discovery of events with a particular
periodicity.
We propose a visualization technique that can represent many sets of
time-stamps. Our visualization technique is based on a circle, like an
analogue clock, and draws all events on this circle. We call our
representation “ChronoView” (see Fig. 4.1). This representation technique
enables us to visualize many events simultaneously. By observing the
positional distribution of events on the circle, we can understand features
of the occurrence time of the event in an intuitive manner. By looking at a
Search WWH ::
Custom Search