The mining of informative rules calls for methods that include different attributes (e.g., weights, quantities, multiple- concepts) suitable for the context of the problem to be analyzed. Previous studies have focused on algorithms that considered individual attributes but ignored the information gain in each rule when the interaction of two or more attributes are taken into account. Motivated by the above, we developed a framework called CrystalBall that supports declarative mining of different rules (i.e., variants) involving several attributes. It eliminates the time and cost of engineering algorithms as practiced in previous studies, and introduces a foundation for cross-variant enhancements. The framework consists of a generic rule mining engine (VI), and a variant description language (VDL) for defining attribute-specific behavior. Be- sides demonstrating the fexibility of the framework, we also discuss the experimental studies, the limitations of the frame- work, as well as future work in the paper.
|Cite as: Ong, K.-L., Ng, W.-K. and Lim, E.-P. (2003). CrystalBall : A Framework for Mining Variants of Association Rules. In Proc. Fourteenth Australasian Database Conference (ADC2003), Adelaide, Australia. CRPIT, 17. Schewe, K.-D. and Zhou, X., Eds. ACS. 85-94. |
(local if available)