Mining Frequent Patterns with Multiple Item Support Thresholds in Tourism Information Databases - Technologies and Applications of Artificial Intelligence - page 96

The procedure for mining the complete set of frequent patterns from compact MIS-
tree is shown in Algorithm 2. The process of mining frequent patterns is described as
follows. Consider the item '122' that has lowest MIS among all items in the compact
MIS tree. It occurs in two branches of compact MIS-tree, <112, 22*, 223, 122: 2>, <112,
21*, 211, 122: 1>. Consider the item '122' as a suffix item , its conditional prefix paths
are <112, 22*, 223: 2> and <112, 21*, 211: 1>, which form the conditional pattern base.
As the compact MIS-tree is constructed, '122' will have MIN value among all items in
its conditional pattern base. Therefore, using MIS value of '122' as the conditional
min_sup, conditional MIS-tree is generated with <112, 22*, 223: 2> and <112, 21*, 211:
1>. The rightmost column of Table 2 lists all the frequent patterns. Similar process is
repeated for other remaining items in the compact MIS-tree to discover the complete set
of frequent patterns. The complete set of frequent patterns is shown in Table 2.

Algorithm 2 CFP-growth++
Input MIS-tree, a set of MIN frequent items F, MIS(i j )
of each item i j in F
Output the complete set of all i j 's conditional frequent
itemsets
Method
1. For each item i j in the header of Tree do
2. Set conditional min_sup, Cmin_sup = MIS(i j )
3. If i j is a frequent item
4. Generate pattern ʲ = i j ∪ ʱ with support = i j .support.

5. Construct ʲ 's conditional pattern base and ʲ 's
conditional MIS-tree T ʲ
6. If T ʲ ≠

ʦ
7. CFPGrowth++( T ʲ , ʲ , Cmin_sup)
8. End for
Procedure CFPGrowth++(Tree, ʱ , Cmin_sup)
1. For each item i j in the header of Tree do
2. Generate pattern ʲ = i j ∪ ʱ with support = i j .support.

3. Construct ʲ 's conditional pattern base and ʲ 's
conditional MIS-tree T ʲ
4. If T ʲ ʦ

5. If T ʲ contains a single path P
6. For each combination ʳ of the nodes in the path P
do
7. Generate pattern ʳ

≠

∪

ʲ with support = minimum
support count of nodes in ʳ .
8. End for
9. Else
10. CFPGrowth++( T ʲ , ʲ , Cminsup).
11. End for