Two new approaches to evaluate association rules
| dc.contributor.author | Delpisheh, Elnaz | |
| dc.contributor.author | University of Lethbridge. Faculty of Arts and Science | |
| dc.contributor.supervisor | Zhang, John Z. | |
| dc.date.accessioned | 2011-08-25T22:45:12Z | |
| dc.date.available | 2011-08-25T22:45:12Z | |
| dc.date.issued | 2010 | |
| dc.degree.level | Masters | |
| dc.description | viii, 85 leaves : ill. ; 29 cm | en_US |
| dc.description.abstract | Data mining aims to discover interesting and unknown patterns in large-volume data. Association rule mining is one of the major data mining tasks, which attempts to find inherent relationships among data items in an application domain, such as supermarket basket analysis. An essential post-process in an association rule mining task is the evaluation of association rules by measures for their interestingness. Different interestingness measures have been proposed and studied. Given an association rule mining task, measures are assessed against a set of user-specified properties. However, in practice, given the subjectivity and inconsistencies in property specifications, it is a non-trivial task to make appropriate measure selections. In this work, we propose two novel approaches to assess interestingness measures. Our first approach utilizes the analytic hierarchy process to capture quantitatively domain-dependent requirements on properties, which are later used in assessing measures. This approach not only eliminates any inconsistencies in an end user’s property specifications through consistency checking but also is invariant to the number of association rules. Our second approach dynamically evaluates association rules according to a composite and collective effect of multiple measures. It interactively snapshots the end user’s domain- dependent requirements in evaluating association rules. In essence, our approach uses neural networks along with back-propagation learning to capture the relative importance of measures in evaluating association rules. Case studies and simulations have been conducted to show the effectiveness of our two approaches. | en_US |
| dc.identifier.uri | https://hdl.handle.net/10133/2530 | |
| dc.language.iso | en_US | en_US |
| dc.publisher | Lethbridge, Alta. : University of Lethbridge, Dept. of Mathematics and Computer Science, c2010 | en_US |
| dc.publisher.department | Department of Mathematics and Computer Science | en_US |
| dc.publisher.faculty | Arts and Science | en_US |
| dc.relation.ispartofseries | Thesis (University of Lethbridge. Faculty of Arts and Science) | en_US |
| dc.subject | Data mining | en_US |
| dc.subject | Association rule mining | en_US |
| dc.subject | Dissertations, Academic | en_US |
| dc.title | Two new approaches to evaluate association rules | en_US |
| dc.type | Thesis | en_US |