Please use this identifier to cite or link to this item: https://ptsldigital.ukm.my/jspui/handle/123456789/513480
Full metadata record
DC FieldValueLanguage
dc.contributor.advisorKhairuddin Omar, Prof. Dr.
dc.contributor.authorWael Ahmad Awad Alzoubi (P35245)
dc.date.accessioned2023-10-16T04:37:08Z-
dc.date.available2023-10-16T04:37:08Z-
dc.date.issued2013-06-21
dc.identifier.otherukmvital:74773
dc.identifier.urihttps://ptsldigital.ukm.my/jspui/handle/123456789/513480-
dc.descriptionA transactional database is created to record all the products purchased by the customers. To focus on the relationships among certain products in the database rather than the huge amount of products in a transaction database, the transaction database can be grouped into different clusters based on different criteria, as the customer needs or transaction records length, the combination of graph based rule mining and clustering concepts is still at a basic level, this is the main motivation to develop a framework for explicit integration of clustering and graph based association rules mining methods so that insights can be drawn from both techniques. The latest research in the area of mining association rules has already focused on developing a method to extract association rules based on graphical representation of the items in the huge transaction datasets. The main problems in graph based association rule mining are the high complexity, the loss of information, and scalability of the graph. Various techniques have been developed to reduce the dataset size without any loss of information, to construct the directed graph, and to choose the most suitable way to represent the generated association rules in order to improve the overall mining of association rules. The aims of this study are to propose: (i) A clustering technique of the transaction dataset that uses the same value of minimum support threshold to generate frequent itemsets of different lengths, (while the previous works change the value of minimum support thresholds during the generation of various lengths of frequent itemsets); (ii) A dataset reduction technique called Dataset reduction based on clustering (DRBC) has been proposed to eliminate infrequent itemsets in successive manner with less loss of information according to the Apriori property; (iii) A graph based association rule technique is proposed, where the association for each frequent itemset is represented by a sub-graph, then all sub-graphs are merged to determine association rules with high confidence and eliminate weak rules, the proposed graph based technique is self-motivated since it builds the association graph in a successive manner. These rules achieve the scalability and reduce the time needed to extract them. Ten synthetic and real transaction datasets are used in the experiments, where twenty different experiments have been conducted. Five experiments are used to evaluate the performance of the proposed clustering technique while ten experiments have been done to compare the proposed graph technique with the triangle counting approach for graph based association rules mining that is considered as a new and efficient technique to prune the association graph in the process of looking for frequent patterns. The final experiments involve evaluation of the proposed dataset reduction algorithm (i.e. DRBC) in terms of the execution time, the rules dimensionality, the number of rules generated, the accuracy and the dimensionality of clusters. The results show that the proposed technique (clustering technique, graph based rule mining technique, and DRBC) outperform the triangle counting approach in all measurements.,Ph.D
dc.language.isoeng
dc.publisherUKM, Bangi
dc.relationFaculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat
dc.rightsUKM
dc.subjectGraph
dc.subjectMining transactional rules
dc.subjectData mining
dc.titleSuccessive graph based approach for mining transactional association rules
dc.typeTheses
dc.format.pages191
dc.identifier.callnoQA76.9.D343.A459 2013 3
dc.identifier.barcode000304
Appears in Collections:Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Files in This Item:
File Description SizeFormat 
ukmvital_74773+Source01+Source010.PDF
  Restricted Access
3.86 MBAdobe PDFThumbnail
View/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.