An Extended Partition Model for Generalized Multidimensional Data

Multidimensional databases have been playing a significant role in the database field. However, some cases may need a more flexible model than the rigid structure of a data cube where all cells in each dimension resolultion combination must be filled. Generally, a data cube may be viewed as a multiple partition over a population. We propose a two-level formalism for modeling multiple partitions and aggregate data assigned to their component sets, based on partition semantics of attributes. On the lower level, partition relations augmented with cumulative attributes represent the partition data. The user level uses a structural graph of attributes based on functional constraints for describing the schema of the database and for expressing query operations in an attribute-oriented way. The model is flexible to deal with cases like sparse or nested data cubes or with complex relationships or hierarchies among partitioning attributes.