OBTAINING AND EVALUATING GENERALIZED ASSOCIATION RULES
Verônica Oliveira de Carvalho, Solange Oliveira Rezende, Mário de Castro · 2007
Generalized association rules are rules that contain some background knowledge giving a more general view of the domain. This knowledge is codified by a taxonomy set over the data set items. Many researches use taxonomies in different data mining steps to obtain generalized rules. So, this work initially presents an approach to obtain generalized association rules in the post-processing data mining step using taxonomies. However, an important issue that has to be explored is the quality of the knowledge expressed by generalized rules, since the objective of the data mining process is to obtain useful and interesting knowledge to support the user’s decisions. In general, what researches do to help the users to select these pieces of knowledge is to reduce the obtained set by pruning some specialized rules using a subjective measure. In this context, this paper also presents a quality analysis of the generalized association rules. The quality of the rules obtained by the proposed approach was evaluated. The experiments show that some knowledge evaluation objective measures are appropriate only when the generalization occurs on one specific side of the rules.