ROC Analysis of Example Weighting in Subgroup Discovery

This paper presents two new ways of example weighting for subgroup discovery. The proposed example weighting schemes are applicable to any subgroup discovery algorithm that uses the weighted covering approach to discover interesting subgroups in data. To show the implications that the new example weighting schemes have on subgroup discovery, they were implemented in the APRIORI-SD algorithm. ROC analysis was then used to study their behavior, and the behavior of APRIORI-SD’s original example weighting scheme, both theoretically and practically, by application on the UK Traffic challenge data set. The findings show that the proposed example weighting schemes are a valid alternative to APRIORI-SD’s original example weighting scheme when the goal is to discover fewer subgroups that are either small and highly accurate or large and less accurate.

• Type: Conference
• Year: 2004
• Where: ROCAI
• Authors: Branko Kavsek, Nada Lavrac, Ljupco Todorovski
• Keywords: Artificial Intelligence, Example Weighting, Example Weighting Scheme, ROCAI 2004, Subgroup Discovery