Database Reference
In-Depth Information
Chapter 9
Conclusions
Uncertain data becomes important and prevalent in many critical applications, such
as sensor networks, location-based services, and health-informatics. Due to the im-
portance of those applications and increasing amounts of uncertain data, effective
and efficient uncertain data analysis has received more and more attentions from
the research community. In this topic, we study ranking queries on uncertain data.
Particularly, we develop three extended uncertain data models based on the state-of-
the-art uncertain data models and propose five novel ranking problems on uncertain
data.
In this chapter, we first summarize the topic, and then discuss some interesting
future directions.
9.1 Summary of the Topic
In this topic, we study ranking queries on uncertain data and make the following
contributions.
•
We develop three extended uncertain data models to fit different application in-
terests: the uncertain data stream model, the probabilistic linkage model, and the
probabilistic road network model.
•
We propose a series of novel
top-k typicality queries
to find the most representa-
tive instances within an uncertain object.
-
A
top-k simple typicality
query returns the top-
k
most typical instances within
an uncertain object. The notion of typicality is defined as the likelihood of an
instance in the object, which is computed using the kernel density estimation
methods.
-
Given two uncertain objects
O
and
S
,a
top-k discriminative typicality query
finds the top-
k
instances in
O
that are typical in
O
but atypical in
S
.
-
A
top-k representative typicality query
finds
k
instances in an uncertain object
that can best represent the distribution of
O
.
207
