A Multi-driven Approach to Improve Data Analytics for Multi-value Dimensions - New Contributions in Information Systems and Technologies - page 114

approach to supply the data sources to determine the DW conceptual schema; and a
final process-driven integration step to model public procurement processes and its
data, based on the requirements and goals identified in the two previous approaches.
The aim was to assure alignment with business goals. By combining the results of the
process-driven, the schema of the conceptual DW can be obtained more easily from a
data-driven approach taking into account the requirements, technological capabilities
and goals identified.
However, in the OOP-DW some of the dimensions are non-strictness, not comply-
ing therefore with the established many-to-one fact-dimension relationship, causing
erroneous data when analysing, for instance all Lots within a specific Procedure,
with more than one Contracting Entity . Such kind of analytic queries can lead
into summarizability problems and double counting of the same fact, especially if we
are using OLAP tools.
To solve the problem it would be enough the place more than one key value in the
fact table, flattening the complexity but this solution is somewhat limited and only
works for a set number of relationships [19]. It will only be a matter of time before we
come across a case where multiple rows of the Contracting Entities collabo-
rate on the same procedure. An alternative is to bridge the fact-dimension relationship
with a bridge table. This design approach is much more flexible in terms of the infor-
mation it can capture, but brings its own set of problems [3]. Another approach is to
use allocations, but in most situations an allocation rule simply does not apply
because some facts cannot be meaningfully allocated.
For the scope of demonstration, Fig. 1 presents a simplified star schema of the
OOP-DW with one fact table Factos and a basic metric defined by the Lot base
price ( PrecoLote) and three dimensions: contracting entity designated as Adju-
dicante , procedure designated as Procedimento and the lot dimension desig-
nated as Lote .
Fig. 1. Star schema to demonstrate the implemented technique