Mining Frequent Patterns with Multiple Item Support Thresholds in Tourism Information Databases - Technologies and Applications of Artificial Intelligence - page 91

to mine both frequent and rare patterns. Experimental results are discussed in section
4. Finally, the conclusion and future work is drawn in section 5.

Fig. 1. An ontology schema for tourism information database

2 Problem Definition

In this section, we discuss rare item problem and extended version of frequent pat-
terns on multiple min_sups.

Table 1. Encoded transaction database

TID Items Items (ordered by MIS

value)

1 {111, 121, 211, 21*} {21*, 111, 211, 121}

2 {111, 121, 221, 22*} {111, 22*, 221, 123}

3 {111, 123, 212, 21*} {21*, 111, 212, 123}

4 {112, 122, 222, 22*} {112, 22*, 222, 122}

5 {112, 122, 211, 21*} {112, 21*, 211, 122}

6 {112, 122, 222, 22*} {112, 22*, 222, 122}

7 {111, 123, 211, 21*} {21*, 111, 211, 123}

8 {112, 123, 211, 21*} {112, 21*, 211, 123}

9 {112, 123, 222, 22*} {112, 22*, 222, 123}

10 {112, 123, 212, 21*} {112, 21*, 212, 123}

2.1 Rare Item Problem

There are mostly non-uniform in nature containing both frequent and rare items in
real world datasets. If the frequencies of items in a database vary widely, the
following issues will be encountered while mining frequent patterns under single
min_sup framework:

1. If min_sup is set too high, the frequent patterns containing rare items will not

be exploited.

2. If min_sup is set too low, it can find frequent patterns that involve both

frequent and rare items. However, it can also cause combinatorial explosion,
generating too many meaningless frequent patterns.