Developing an efficient mechanism for mining association rules in small large transactional databases

Dissertant

al-Sharman, Husam Muhammad

Thesis advisor

al-Bahadili, Husayn

Comitee Members

al-Shammari, Husayn Hadi Uwayyid
al-Shaykh, Asim A. R.
al-Lahham, Muhammad Ismail Abd al-Rasul

University

Arab Academy for Financial and Banking Sciences

Faculty

The Faculty of Information Systems and Technology

Department

Computer information systems

University Country

Jordan

Degree

Ph.D.

Degree Date

2012

English Abstract

This thesis presents a description of a newly developed high performance association rule mining algorithm that basically combines the well-known Apriori association rule mining algorithm with a compact mechanism, namely, the Dynamic Compact Grouping Item sets (DCGI).

DCGI is specially designed for optimizing online and very large transactional databases.

Therefore, this new algorithm is called Online Dynamic Grouping Association Rule Mining (ODGARM) algorithm.

ODGARM has a number of advantages over existing algorithms, such as: the first to provide efficient and accurate online dynamic grouping association rule mining on transactional databases of various sizes, allow generating frequent patterns quickly by skipping the repetitive database scan and reducing a great amount of time per database scan; reduce the search space (candidate's item sets) and discover only those rules which can be interesting for the user; reduce the memory requirements to store a huge number of useless candidates; and enable good diversification of the search space and in many cases decreases the time needed for the algorithm to generate a set of large frequent item sets.

The main corner stone of ODGARM are: inclusion of DCGI as a compact pre-processing mechanism, penalizing rules with few item sets using weight thresholds, inclusion of a new mechanism supported with developed mathematical model to prevent eliminating good item sets that might be part of the best solutions we are looking for, avoid one-at-a-time transaction processing and retrieval from the database, and avoid spending time improving low starting quality solution.

In order to evaluate and compare the performance of ODGARM with other algorithms in the literature, four scenarios are simulated using synthetic and real datasets.

The first three scenarios investigate the effect of input parameters, such as: minimum support, number of items in item set, and number of database transactions.

The forth scenario presents a comparison between the performance of ODGARM against the performance of other algorithms found in the literature (e.g., Apriori, Apriori-Tad, Apriori Hybrid, FP-Growth, and Partition).

Main Subjects

Information Technology and Computer Science

Topics

• Data processing
• Databases
• Data mining

No. of Pages

84

Table of Contents

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-306659