An Efficient Exception Mining Algorithm in Multi-dimensional Data Cube

In the process of OLAP analysis based on multi-dimensional data, analysts are often involved in large-scale data cube, which results users cannot find the interest information efficiently. To overcome this problem, some exceptions mining or exceptions-based methods were proposed. In this paper, a new regression-based definition of exception is proposed, threshold exception, and following which an exception mining algorithm is proposed to help users find the exceptions in the data cells effectively using regression parameters. This method estimates the data as exception by comparing its normalized residual to the thresholds user gave. Performance study shows that the method is practical and effective.