Mining Frequent Patterns with Multiple Item Support Thresholds in Tourism Information Databases - Technologies and Applications of Artificial Intelligence - page 94

linked in sequence via such node links. After scanning all the transactions in Table 1,
we will get the initial MIS-tree shown in Fig. 2.
To decrease the search space, the compact MIS-tree is generated by pruning items
with supports less than MIN value and merging nodes with the same item name. For
example, we remove the nodes with items, 121 and 221, and the complete and com-
pact MIS-tree is shown in Fig. 3.
Fig. 3. Compact MIS-tree
Input: a transaction database DB and an ontology
Output: MIS-tree
Method:
1. Create the root of a MiS-tree, T, and label it as null
2. For each Transaction, t n , in encoded TDB do
3. Sort all items in t n according to their MIS(i)in
deccending order
4. Count the support values of any item i, denoted as
Sup(i) in t n .
5. Insertion(t n [p|P], T), where p is the first element
and P is the remaining list.
6. End for
7. For each item f in unfrequent itemset F do where F is
the set of items with support smaller than MIN(F)
8. Delete the entry in the header table with item_name =
f
9. Pruning(T, f)
10.End for
11.Get MIN frequent item header table
12.Merge(T)
Procedure Insertion(t n [p|P], T)
1. While P is not empty