Promotion Analysis Based on Fitting in Multi-Level Space

Abstract

This paper addresses an interesting problem, which is to find top-k L-Regions of a given object in a fact table with multiple dimensions and each dimension has one hierarchy with multiple levels. L-Region is a variant of combination of members from each dimension. To tackle the problem of the large solution space, we propose a materialized strategy called LR-Cube, which balances storage and time overhead by model fitting. The LRAPQuery designed in this paper is a query guaranteeing completeness and uniqueness in checking LRegions. The experiments show that our proposed method is both feasible and of high efficiency.