The Incremental Mining Of Constrained Cube Gradients

The mining of cube gradients is an extension of traditional association rules mining in data cube and has broad applications. In this paper, we consider the problem of mining constrained cube gradients for partially materialized data cubes. Its purpose is to extract interesting gradient-probe cell pairs from partially materialized cubes while adding or deleting cells. Instead of directly searching the new data cubes from scratch, an incremental mining algorithm IncA is presented, which sufficiently uses the mined cube gradients from old data cubes. In our algorithms, the condensed cube structure is used to reduce the sizes of materialized cubes. Moreover, some efficient methods are presented in IncA to optimize the comparison process of cell pairs. The performance studies show the incremental mining algorithm IncA is more efficient and scalable than the directed mining algorithm DA with different constraints and sizes of materialized data cubes.