Incremental constraint class association rule mining of student performance dataset

摘要

In Associative Classification (AC), Class Association Rules are generally used in the process of classification in the field of medicine, education, business and so on. AC generates huge number of association rules which consumes memory and mining time. Since users are interested in only useful and interesting class association rules, constraints are introduced in the generation of Class Association Rules (CAR) to reduce the memory and mining time. Only those rules with at least one constraint items are used in the process of associative classification to improve accuracy. The major problem in associative classification is dynamic dataset. The proposed framework focuses on incremental data which is used in most of the real time applications. It makes use of Incremental Constraint Class Association Rule (ICCAR) Algorithm by constructing Incremental Constraint Class Rule (ICCR) Tree to generate the Constraint Class Association Rules. In case of incremental data, the ICCR Tree is updated for the new set of records without re-scanning the dataset. A metric namely, safety threshold is used to determine whether the dataset to be re-scanned or not. Also, since the nodes of the ICCR Tree is made to store the differences of the object identifiers, it consumes less memory when compared to existing associative classifiers.