Association rules are a useful concept in data mining with the goal of summa- rizing the strong patterns that exist in data. We have identified several issues in mining association rules and addressed them in three main areas. The first area we explore is standardized interestingness measures. Different interestingness measures exist on different ranges, and interpreting them can be subtly problematic. We standardize several interestingness measures and show how these are useful to consider in association rule mining in three examples. A second area we address is incomplete transactions. By applying statistical methods in new ways to association rules, we provide a more comprehensive means of analyzing incomplete transactions. We also describe how to find families of distributions for interestingness measure values when transactions are incomplete. Finally, we address the common result of mining: a plethora of association rules. Unlike methods which attempt to reduce the number of resulting rules, we harness this large quantity to find a higher-level set of patterns. / NSERC Discovery Grant and OMRI Early Researcher Award
Identifer | oai:union.ndltd.org:LACETR/oai:collectionscanada.gc.ca:OGU.10214/7250 |
Date | 21 June 2013 |
Creators | Shaikh, Mateen |
Contributors | McNicholas, Paul |
Source Sets | Library and Archives Canada ETDs Repository / Centre d'archives des thèses électroniques de Bibliothèque et Archives Canada |
Language | English |
Detected Language | English |
Type | Thesis |
Rights | Attribution-NonCommercial-NoDerivs 2.5 Canada, Attribution-NonCommercial-NoDerivs 2.5 Canada, Attribution-NonCommercial-NoDerivs 2.5 Canada, http://creativecommons.org/licenses/by-nc-nd/2.5/ca/, http://creativecommons.org/licenses/by-nc-nd/2.5/ca/, http://creativecommons.org/licenses/by-nc-nd/2.5/ca/ |
Page generated in 0.0019 seconds