Subspace Discovery for Promotion: A Cell Clustering Approach

The promotion analysis problem has been proposed in [16], where ranking-based promotion query processing techniques are studied to effectively and efficiently promote a given object, such as a product, by exploring ranked answers. To be more specific, in a multidimensional data set, our goal is to discover interesting subspaces in which the object is ranked high. In this paper, we extend the previously proposed promotion cube techniques and develop a cell clustering approach that is able to further achieve better tradeoff between offline materialization and online query processing. We formally formulate our problem and present a solution to it. Our empirical evaluation on both synthetic and real data sets show that the proposed technique can greatly speedup query processing with respect to baseline implementations.