Database Reference
In-Depth Information
{
f
}:2.1
α
2
+2.1
{
g
}:1.8
α
2
+2.1
α
+2.7
{
h
}:1.5
α
2
+1.5
{
i
}:1.0
{
f,g
}:1.26
α
2
+1.89
{
f,h
}:1.05
α
2
+1.05
{
g,h
}:1.35
α
2
+1.35
{
g,i
}:0.90
{
f,g,h
}:0.945
α
2
+0.945
Fig. 14.10
The TUF-stream structure for streaming uncertain data
D
3
Fig. 14.11
The LUF-stream
structure for streaming
uncertain data
D
3
{
g
}:6.6
{
h
}:3.0
{
i
}:1.0
{
f
}:4.2
{
f,g
}:3.15
{
f,h
}:2.10
{
g,h
}:2.70
{
g,i
}:0.90
{
f,g,h
}:1.890
8.4
LUF-streaming for the Landmark Model
Moreover, Leung et al. [
41
] extended the TUF-streaming algorithm to become the
LUF-streaming
algorithm, which mines frequent patterns from streaming uncertain
data in a fashion similar to the UF-streaming and TUF-streaming algorithms, except
that LUF-streaming uses the
landmark model
. When using the landmark model, the
corresponding
LUF-stream
structure (i) captures all “frequent” patterns mined from
all batches of streaming uncertain data generated from a landmark to the present time
and (ii) treats all batches with the same importance. See Fig.
14.11
.
8.5
Hyperlinked Structure-Based Streaming Uncertain Frequent
Pattern Mining
So far, we have described notable exact and approximate
tree-based
streaming un-
certain frequent pattern mining algorithms. Besides them, there are also hyperlinked
structure based algorithms. For instance, Nadungodage et al. [
46
] proposed two
false positive-oriented algorithms called
UHS-Stream
and
TFUHS-Stream
. The
UHS-Stream algorithm applies uncertain hyperlinked structure stream mining for
finding all the frequent patterns seen up to the current moment (i.e., with the land-
mark model). Similarly, the TFUHS-Stream algorithm applies uncertain hyperlinked
structure stream mining, but it uses the time-fading model. Hence, the TFUHS-
Stream algorithm puts heavier weights on the recent transactions than historical data
in the stream.
