Enhancing classification accuracy with frequent positive and negative rules

Anis Suhailis Abdul Kadir, Azuraliza Abu Bakar, Abdul Razak Hamdan

Research output: Contribution to journalArticle

Abstract

Associative classification has been proven to be more accurate than the state-of-the-art classification algorithms, such as C4.5. The rules known as class association rules (CARs) are used to build the classifier. Initially, positive association rules were generated to build associative classifiers. However, more recently, negative association rules have been recognized for their ability to enhance associative classification accuracy. Literature shows that the knowledge obtained from negative association rules is considered unique and surprising compared to the positive association rules. We propose to mine the quality of negative association rule from frequent positive and negative (FPN) itemset approach. The FPN approach will be embedded in an Apriori algorithm for mining negative association rules and later, integrated with a CBA algorithm for the construction of the classifier. This approach presents challenges in the search space and selecting quality CARs in order to enhance the classification accuracy. An experiment was conducted with UCI datasets to evaluate the classifier's performance and the results demonstrated that the FPN managed to produce competitive classifier.

Original languageEnglish
Pages (from-to)699-713
Number of pages15
JournalJournal of Theoretical and Applied Information Technology
Volume66
Issue number3
Publication statusPublished - 31 Aug 2014

Fingerprint

Association rules
Association Rules
Negative Association
Classifiers
Classifier
Apriori Algorithm
Classification Algorithm
Search Space
Mining
Evaluate

Keywords

  • Associative classification
  • Negative association rule

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Enhancing classification accuracy with frequent positive and negative rules. / Abdul Kadir, Anis Suhailis; Abu Bakar, Azuraliza; Hamdan, Abdul Razak.

In: Journal of Theoretical and Applied Information Technology, Vol. 66, No. 3, 31.08.2014, p. 699-713.

Research output: Contribution to journalArticle

@article{076861c5e3af4566ae5c122f26208bc1,
title = "Enhancing classification accuracy with frequent positive and negative rules",
abstract = "Associative classification has been proven to be more accurate than the state-of-the-art classification algorithms, such as C4.5. The rules known as class association rules (CARs) are used to build the classifier. Initially, positive association rules were generated to build associative classifiers. However, more recently, negative association rules have been recognized for their ability to enhance associative classification accuracy. Literature shows that the knowledge obtained from negative association rules is considered unique and surprising compared to the positive association rules. We propose to mine the quality of negative association rule from frequent positive and negative (FPN) itemset approach. The FPN approach will be embedded in an Apriori algorithm for mining negative association rules and later, integrated with a CBA algorithm for the construction of the classifier. This approach presents challenges in the search space and selecting quality CARs in order to enhance the classification accuracy. An experiment was conducted with UCI datasets to evaluate the classifier's performance and the results demonstrated that the FPN managed to produce competitive classifier.",
keywords = "Associative classification, Negative association rule",
author = "{Abdul Kadir}, {Anis Suhailis} and {Abu Bakar}, Azuraliza and Hamdan, {Abdul Razak}",
year = "2014",
month = "8",
day = "31",
language = "English",
volume = "66",
pages = "699--713",
journal = "Journal of Theoretical and Applied Information Technology",
issn = "1992-8645",
publisher = "Asian Research Publishing Network (ARPN)",
number = "3",

}

TY - JOUR

T1 - Enhancing classification accuracy with frequent positive and negative rules

AU - Abdul Kadir, Anis Suhailis

AU - Abu Bakar, Azuraliza

AU - Hamdan, Abdul Razak

PY - 2014/8/31

Y1 - 2014/8/31

N2 - Associative classification has been proven to be more accurate than the state-of-the-art classification algorithms, such as C4.5. The rules known as class association rules (CARs) are used to build the classifier. Initially, positive association rules were generated to build associative classifiers. However, more recently, negative association rules have been recognized for their ability to enhance associative classification accuracy. Literature shows that the knowledge obtained from negative association rules is considered unique and surprising compared to the positive association rules. We propose to mine the quality of negative association rule from frequent positive and negative (FPN) itemset approach. The FPN approach will be embedded in an Apriori algorithm for mining negative association rules and later, integrated with a CBA algorithm for the construction of the classifier. This approach presents challenges in the search space and selecting quality CARs in order to enhance the classification accuracy. An experiment was conducted with UCI datasets to evaluate the classifier's performance and the results demonstrated that the FPN managed to produce competitive classifier.

AB - Associative classification has been proven to be more accurate than the state-of-the-art classification algorithms, such as C4.5. The rules known as class association rules (CARs) are used to build the classifier. Initially, positive association rules were generated to build associative classifiers. However, more recently, negative association rules have been recognized for their ability to enhance associative classification accuracy. Literature shows that the knowledge obtained from negative association rules is considered unique and surprising compared to the positive association rules. We propose to mine the quality of negative association rule from frequent positive and negative (FPN) itemset approach. The FPN approach will be embedded in an Apriori algorithm for mining negative association rules and later, integrated with a CBA algorithm for the construction of the classifier. This approach presents challenges in the search space and selecting quality CARs in order to enhance the classification accuracy. An experiment was conducted with UCI datasets to evaluate the classifier's performance and the results demonstrated that the FPN managed to produce competitive classifier.

KW - Associative classification

KW - Negative association rule

UR - http://www.scopus.com/inward/record.url?scp=84906890155&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84906890155&partnerID=8YFLogxK

M3 - Article

VL - 66

SP - 699

EP - 713

JO - Journal of Theoretical and Applied Information Technology

JF - Journal of Theoretical and Applied Information Technology

SN - 1992-8645

IS - 3

ER -