Database Reference
In-Depth Information
event is defined to occur at a point of time, usually called as a
point
event
; while a composite event can always span a period of time, i.e.,
an
interval event
[72]. Any dimension of attributes of an event can
be either certain or uncertain. An event with one or multiple uncertain
attributes is an
uncertain event
or a
probabilistic event
[53]; other-
wise, it is a
certain event
. If an event cannot detect its occurrence by
itself unless it either gets expired or is explicitly queried, we name it as a
non-spontaneous event
[73]. RFID applications and sensor applica-
tions can generate non-spontaneous events such as negated events and
temporal constrained events. Such non-spontaneous events pose new
challenges for event processing.
It is common that real-world events are associated with time and spa-
tial or location dimensions, which mirror the most common inquiries
about such events, i.e.,
when
and
where
. However, events can contain
more information than these two well-known dimensions entail. Orig-
inally, other semantic properties such as genealogy, identification and
others can describe partitional information of an event. All these prop-
erties of an event can be viewed as its context, namely
event context
,
and event context can be temporal, spatial, semantic, or even social.
Contexts of events can significantly affect the semantics for event pro-
cessing, and it is critical to identify the context and the type of context
to process events effectively and semantically.
1.2 Event Processing
Vast amounts transaction data and monitoring data can be constantly
generatedaseventstreams,whichhavetobefullyprocessedtosupport
automated business decisions or time-critical actions. Basically, events
cannot be entirely foreseen [62], and we cannot predict whether a critical
event will happen, or when it will happen. In reality, what we can do is to
ensure that the interested events can be detected in a real-time or quasi-
real-time manner; this is the main purpose of
event processing
.Thus
timeliness is among the top priorities in event processing applications.
Generally speaking, event processing can be broadly defined to be any
computing that performs operations on events, including reading, creat-
ing, transforming and deleting events [28]. The main idea of event pro-
cessing is to process events to gather meaningful or valuable information
and then deriving actions from them. The main functional capabilities
required by event processing applications include data filtering, aggrega-
tion, transformation, pattern detection, pattern discovery and pattern
prediction. Non-functional requirements include performance, response
time and throughput.
