Abstract:
Frequent-regular itemsets/petterns mining has been explored and proposed
to find interesting itemsets in a database based on their own occurrence behavior.
Traditionally, an itemset can be identified as interesting by considering only frequency
and regularity of an itemset occurred in the database. However, itemsets can have
different degree of importance which traditional approach may affect the missing of
important/interesting knowledge in real-world applications. In this thesis, we introduce
approaches on mining weighted-frequent-regular itemsets (also called mining WFRIs).
in which the first approach is called Weighted-Frequent-Regular Itemsets Miner (WFRIM)
by using FP-tree like structure named WFRI-treeto maintaincandidate itemsets during
mining process and using WFRIM-growthtechnique to mine WFRIs. To improve WFRIM,
the second approach is proposed called Weighted-Frequent-Regular Itemset Miner
using Interval Word Segment structure (WFRIM-IWS). The dynamic bit vector is utilized
for maintaining occurrence information of each itemsetnamed interval word segments
structure (IWS). The both approaches apply the concept of overestimated weightedfrequency and global/local maximum weights to early prune search spaceand reduce
computational time. From experimental results on synthetic and real datasets, the
both approaches can exhibit to discover weighted-frequent-regular itemsetsefficiently.
In addition, WFRIM-IWS outperforms WFRIM in the terms of computational time and
memory consumptio