Irrelevant Feature and Rule Removal for Structural Associative Classification Using Structure-Preserving Flat Representation

摘要: Practical applications of association rule mining often suffer from overwhelming number rules that are generated, many which not interesting or useful for the application in question. Removing irrelevant features and/or comprised can significantly improve overall performance. Many statistical and constraint based measures used to discard unnecessary when vectorial tabular data is In contrast, use such limited tree-structured domain, due structural aspects easily incorporated. this chapter, we explore a feature subset selection measure as well common interestingness via recently proposed structure-preserving flat representation XML. A prior generation. Once initial set obtained, determined those attributes be statistically significant classification task. The experiments performed using real world web access trees property management dataset. results indicate where dataset has more standard structure large insignificant will discarded accuracy increase. However, tree instances vary greatly terms label distribution among nodes, while removed increases, there reduction coverage rate set.