Associative classification technique based on incremental mining

Dissertant

al-Nababteh, Muhammad Husayn Abd al-Hamid

Thesis advisor

al-Shalabi, Riyad

Comitee Members

al-Shaykh, Asim A. R.
al-Sarayirah, Bashshar
al-Muidi, Hasan

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

2011

English Abstract

With the advance in storage and data collection devices, organization enabled to collect and store vast amount of data, consequently, extracting information become big challenging.

Data mining is the process of automatically discovering valuable information from huge dataset.

Associative classification (AC) is an approach in data mining that uses association rule to build classification systems that are easy to interpret by end-user.

Previous studies proposed that associative classification has high classification accuracy and strong flexibility at handling unstructured data.

However, it still suffers from lack of their ability to deal with the update of the original dataset.

When different data operations (adding, deleting, updating) are applied against certain training data set, current AC algorithms must scan the complete training data set again to update the results (classifier) in order to reflect change caused by such operations.

This thesis deals with incremental data (i.e.

data insertion) issue within the incremental mining in AC.

A new associative classification method developed called ACIM, i.e., Associative Classification based on Incremental Mining.

Particularly, a known AC algorithm called Classification based on Association Algorithm (CBA) modified to treat one aspect of the incremental mining problem which is data insertion operation.

Our algorithm doesn’t rebuild the classifier from scratch after each modification which either use the same classifier to classify unseen objects, or create new classifier by utilizing the current classifier rules and incremental data.

Experimental results against six data sets from UCI machine learning data repository showed that the ACIM algorithm reduces the computational time if compared to CBA 2-7 times and almost derives the same accuracy of them.

Main Subjects

Information Technology and Computer Science

Topics

• Data processing
• Data mining

No. of Pages

104

Table of Contents

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-306731