Information Technology Reference
In-Depth Information
2. If
T
has a child
N
such that
N
.item-no=
p
, then increment
N
.count by 1.
3. If
P
is not empty, then call insert_tree(
P, N
) recursively.
TID
Item
root
1
abcefo
2
acg
0:8
2:2
header
fitems
a:8
3
ei
0
4
1:8
acdeg
c:8
e:8
1
2
5
acegl
g:5
3
2:6
3:1
5:1
6
ej
b:2
d:2
4
5
7
abcefp
8
acd
3:4
4:2
f:2
6
9
acegm
5:1
10
acegn
6:2
a)
b)
Fig. 12.3. CFP-tree (minsup=20%)
After the construction of CFP-tree, we should change the sn-link from
sibling-link to next-link and reverse the cp-link. The processing procedure is as
follows: Traverse the tree from the root. Add current node CN to the link of
header[CN.item-no] as the last node. If CN has no child or all of its children and
siblings have been processed then let CN.cp-link=CN's parent, else process its
children and its siblings recursively.
Figure 12.3 (a) shows an example of a database and Figure 12.3 (b) is the
CFP-tree for that database.
Here give a CFP-tree mining algorithm based on constrained subtree and
array-based technique(Grahne, 2003). The following is the pseudocode of
CFP-tree mining algorithm.
Algorithm 12.5 CFP-tree mining algorithm
Input: A CFP-tree T
Output: The complete set of FI's corresponding to T
Method:
1. patlen=1;
Search WWH ::




Custom Search