Multidimensional Query Reformulation with Measure Decomposition

Measurement and comparison of performances in networked organisations is particularly critical because of heterogeneity and sparsity of data. In particular, each organization is autonomous in the definitions of which measures to use and their calculation formulas, i.e. the mathematical expressions stating how a measure is calculated from others. Hence, full integration of data marts requires a reconciliation among such heterogeneous definitions in order to support evaluation of cross-organizations performances and to produce meaningful comparisons. To address this issue, this paper proposes (1) an extension of the traditional multidimensional model by taking into account the explicit representation of the semantics for measure formulas, and, on the top of this model, (2) a novel query reformulation approach for a scenario of federated data warehouses. The approach exploits both aggregation and, unlike traditional approaches, measure decomposition through the calculation of measure formulas. This extends usual features of query rewriting based on views, allowing to overcome heterogeneities at measure level among data mart schemas and enabling meaningful comparisons among values of different autonomous data marts. A formalization of the rewriting algorithm is proposed, together with a computational analysis, proofs of correctness and termination, and an evaluation of effectiveness that shows how the approach can lead to a significant increase in the capability of integrating indicators to answer queries in a federated scenario.