A Multi-driven Approach to Improve Data Analytics for Multi-value Dimensions - New Contributions in Information Systems and Technologies - page 113

• Participating Entity , interested entities can respond individually or

in a consortium of entities to each Lot with a distinct Proposal . It is also
possible that the same participant entity presents more than one project pro-
posal to the same Procedure , which will also derive into a non-strictness
situation.

The fact grain is defined for the most granular price value, which corresponds to
the price of a Proposal presented by a Participant entity (or consortium) to a
specific Lot within a Procedure . This means that the base price of a Procedure
corresponds to the SUM of the base price related to each Lot within that Proce-
dure . Consequently, there is a high probability to generate erroneous data when
analysing results correlating data form the Proposal dimension with data from
the Contracting Entity dimension.
To solve these kinds of problems, most of the existing approaches present a set of
informal guidelines to transform complex multidimensional structures into multidi-
mensional structures enforcing summarizability. However, such approaches require
manual decisions and a lot of expertise, which reduces their applicability [17]. The
approach prosed in this paper enables designers to freely specify the DW schema at
the conceptual level because all the computation related to summarizability issues is
automatically handled at the physical level.

4 Technology-driven Approach to Overcome Summarizability

Problems

In the literature it is mentioned that data models obtained from a single approach are
usually incomplete, which cannot obtain satisfaction and trust of organizations and
individuals simultaneously [4, 9]. Indeed the combination of different approaches to
the design of the DW schema allowed us to obtain the right conceptual model of the
DW, i.e., a data model more aligned with organizational and business-user's needs. In
this section, the emphasis will be given to the technique (i.e., technology-driven ap-
proach) used to overcome a critical limitation of some OLAP tools - capability to deal
with multi-value dimensions, without using a bridge table, as it is the case of the Pen-
taho tool. In the existing literature the specific method of handling non-strictness rela-
tionships without increasing design complexity (e.g., through bridging tables or
weighting factors) is also rarely addressed [12, 13].
The goal in designing the DW model is to keep it simple to understand, simple to
load with operational data and as fast as possible to query. This means that, the flatter
the dimensional model, the better it is for business-users [18]. The more complex the
model, the more complex will be to empower business-users to perform their own
queries and analyses. To reach such level of expectations, the DW design must bal-
ance elegance in conceptual design with understandability, usability and performance.
The design of the OOP-DW included a top-down goal-driven and a user-driven ap-
proach performed in parallel to model high level aspects related to public procure-
ment and detailed requirements specification; followed by a bottom-up data-driven