Significant Gradients Mining Based on Data Cube Computation

Gradient analysis is an important data analysis task in data warehousing and online analytical processing, which has played an important role in the application of decision support. This paper considers a novel type of gradient analysis, significant gradient analysis. Significant gradient analysis is expressve, capable of capturing trends in data and answering "what-if" questions. The problem of mining significant gradients is challenging since the significant gradients can be widely scattered in the cube lattice, and do not present any monotonicity. To tackle the problem and develop techniques to speed up the search the state-of-the-art cube computation algorithm is extemded. An extensive perform- ance study is reported to illustrate the effect of the approach.