A hybrid strategy for minimizing association rules

مقدم أطروحة جامعية

Ali, Sayf Muhammad

مشرف أطروحة جامعية

al-Rababiah, Mamun S.
al-Hammami, Ala Husayn

أعضاء اللجنة

Shatnawi, Umar Ali
al-Batayinah, Khalid
Hanandeh, Isam Said S.

الجامعة

جامعة آل البيت

الكلية

كلية الأمير الحسين بن عبد الله لتكنولوجيا المعلومات

القسم الأكاديمي

قسم علوم الحاسوب

دولة الجامعة

الأردن

الدرجة العلمية

ماجستير

تاريخ الدرجة العلمية

2010

الملخص الإنجليزي

Associative classification is an integrating association rules mining and classification.

Associative Classification discovered patterns for useful information are important and in great demand in science, engineering and business.

Today, effective associative classification methods have been developed and widely used in various challenging industrial and business applications.

These methods attempt to provide significant and useful information for decision makers.

Paradoxically, associative classification itself can produce such huge amounts of rules that poses a knowledge management problem: to tackle thousands or even more rules discovered in a dataset.

In order to solve the rules number problem, in this study associative classification process involves two stages.

First: the data is preprocess to improve the classification ability and the efficiency of the classification process.

Second: by using concept of sample itemset matching, the hybrid system of data post-analysis is proposed.

The hybrid system includes data pruning, data clustering and data summarization to support effective analysis and interpretation of the discovered rules.

In data pruning the frequent and redundant itemsets are prune and data clustering using to grouping those itemsets results from data pruning into similar clusters.

Finally, data summarization using to prune itemsets in each clusters those results from data clustering.

In the Hybrid Strategy approach, we use the composite criteria of minimum support and minimum confidence as the rule weight to indicate the significance of the rule.

Through our study, we find it is important to find a good combination of these two rule interestingness threshold values. By using datasets from UCI Repository, we compared the Hybrid Strategy results with some well-known associative classification algorithms such as CMAR, CPAR, TFPC and Classification by Bagged Consistent Itemset Rules.

The results that achieved by Hybrid Strategy not only give better classification accuracy but also have better results of rules number.

Hybrid Strategy generates a much smaller set of high-quality predictive rules with high classification accuracy compare with these algorithms.

التخصصات الرئيسية

تكنولوجيا المعلومات وعلم الحاسوب

الموضوعات

• الإدارة
• قواعد البيانات
• التنقيب في البيانات

عدد الصفحات

72

قائمة المحتويات

• APA
• MLA
• AMA

لغة النص

الإنجليزية

نوع البيانات

رسائل جامعية

رقم السجل

BIM-307564