Efficient computation of multi-feature data cubes

A Multi-Feature Cube (MF-Cube) query is a complex-data-mining query based on data cubes, which computes the dependent complex aggregates at multiple granularities. Existing computations designed for simple data cube queries can be used to compute distributive and algebraic MF-Cubes queries. In this paper we propose an efficient computation of holistic MF-Cubes queries. This method computes holistic MF-Cubes with PDAP (Part Distributive Aggregate Property). The efficiency is gained by using dynamic subset data selection strategy (Iceberg query technique) to reduce the size of materialized data cube. Also for efficiency, this approach adopts the chunk-based caching technique to reuse the output of previous queries. We experimentally evaluate our algorithm using synthetic and real-world datasets, and demonstrate that our approach delivers up to about twice the performance of traditional computations.