Database Reference
In-Depth Information
26.15
2ε
Raw Data
PMC-MR
26.1
26.05
26
25.95
25.9
04.22.08 20:09
04.23.08 00:57
Date & Time (mm/dd/yy hours:min)
Figure 2.12.
Poor Man's Compression - MidRange (PMC-MR).
where
is the maximum allowed approximation error according to the
L
∞
norm. Also, for PMC-Mean and PMC-MR the sensor values in a
segment
g
k
should satisfy the following condition:
max
1
≤p≤i
k
v
i
k−
1
+
p
−
min
1
≤p≤i
k
v
i
k−
1
+
p
≤
2
.
(2.14)
Furthermore, for PMC-Mean, the approximation value for the segment
g
k
is given by the mean value of the sensor values in segment
g
k
.But,
for PMC-MR it is given as follows:
max
1
≤p≤i
k
v
i
k−
1
+
p
−
min
1
≤p≤i
k
v
i
k−
1
+
p
.
2
The data segmentation approach for PMC-MR is illustrated in
Figure
Moreover, the linear filter [34] is a simple piecewise linear approxi-
mation technique in which the sensor values are approximated by a line
connecting the first and second point of the segment. When a new data
tuple cannot be approximated by this line with the specified error bound,
a new segment is started. In [20], two new piecewise linear approxima-
tion models were proposed, namely
Swing
and
Slide
, that achieve much
higher compression compared to the cache and linear filters. We briefly
discuss the swing and slide filters below.
5.3.1 Swing and Slide Filters.
The swing filter is capable of
approximating multi-dimensional data. But, for simplicity, we describe
