Combining Data-Driven and User-Driven Evaluation Measures to Identify Interesting Rules
Solange Oliveira Rezende (University of São Paulo, Brazil), Edson Augusto Melanda (Federal University of São Carlos, Brazil), Magaly Lika Fujimoto (University of São Paulo, Brazil), Roberta Akemi Sinoara (University of São Paulo, Brazil) and Veronica Oliveira de Carvalho (University of Oeste Paulista, Brazil)
Copyright: © 2009
Association rule mining is a data mining task that is applied in several real problems. However, due to the huge number of association rules that can be generated, the knowledge post-processing phase becomes very complex and challenging. There are several evaluation measures that can be used in this phase to assist users in finding interesting rules. These measures, which can be divided into data-driven (or objective measures) and user-driven (or subjective measures), are first discussed and then analyzed for their pros and cons. A new methodology that combines them, aiming to use the advantages of each kind of measure and to make user’s participation easier, is presented. In this way, data-driven measures can be used to select some potentially interesting rules for the user’s evaluation. These rules and the knowledge obtained during the evaluation can be used to calculate user-driven measures, which are used to aid the user in identifying interesting rules. In order to identify interesting rules that use our methodology, an approach is described, as well as an exploratory environment and a case study to show that the proposed methodology is feasible. Interesting results were obtained. In the end of the chapter tendencies related to the subject are discussed.
This section presents some challenges related to the use of association rules, the objective and subjective aspects of the interestingness, objective and subjective measures for the evaluation of association rules and some environments that support the identification of interesting association rules.