Nowadays, the data warehouse is recognized as the essential component of decision support systems since it ensures the best response to the decision problems of different functional areas of an organization. However, designing and building a data warehouse remain a very complex task, difficult to accomplish. This complexity is mainly due to the absence of technics and methods that are recognized in the field. Thus, the present paper identifies different rules for designing a data warehouse from relational data and introduces a new method that aims to automate this process using MDA techniques and XML.

Figures - uploaded by Nawfal el Moukhi

Author content

All figure content in this area was uploaded by Nawfal el Moukhi

Content may be subject to copyright.

ResearchGate Logo

Discover the world's research

  • 20+ million members
  • 135+ million publications
  • 700k+ research projects

Join for free

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 1

Towards a new automatic data warehouse design

method

Ver s une nou ve lle th od e auto ma ti que de co ncept io n des en tre pôt s de do nnée s

Nawfal El Moukhi

MISC Laboratory, Faculty of Sciences, Ibn Tofail University, Kenitra, Morocco

elmoukhi.nawfal@gmail.com

Ikram El Azami

MISC Laboratory, Faculty of Sciences, Ibn Tofail University, Kenitra, Morocco

akram_elazami@yahoo.fr

Abdelaaziz Mouloudi

MISC Laboratory, Faculty of Sciences, Ibn Tofail University, Kenitra, Morocco

mouloudi_aziz@hotmail.com

Abdelali Elmounadi

LASTIMI, Mohammadia School of Engineers, Mohammed V University , Rabat, Morocco

a.elmounadi@gmail.com

Résumé

Les entrepôts de données sont actuellement reconnus comme étant un composant essentiel

des systèmes d'aide à la cision dans la mesure où ils offrent la meilleure réponse aux

problèmes de prise de cision des différents domaine s fonctionnels des organisations.

Cependant, la conception et la construction des entrepôts de données demeurent une tâche

très complexe, difficile à accomplir. Cette complexité est ess entiellement la conséquence de

l'absence de techniques et de méthodes reconnues dans le domaine. Dans ce contexte, cet

article identifie différentes règles pour la conception des entrepôts de données à partir de

données relationnelles et introduit une nouvelle méthode pour l'automatisation de ce

processus en se fondant sur les technologies MDA et XML.

Abstract

Nowadays, the data warehouse is recognized as the essential component of decision support

systems since it ensures the best response to the decision problems of different functional

areas of an organization. However, designing and building a data warehouse remain a very

complex task, dif ficult to accomplish. This complexity is mainly due to the absence of

technics and methods that are recognized in the field. Thus, the present paper identifies

different rules for designing a data warehouse from relational data and introduces a new

method that aims to automate this process using MDA techniques and XML.

Mots clés

Entrepôt de données, modèle relationnel, modèle multidimensionnel,

conception d'entrepôts de données, Architecture orientée modèle

Keywords

Data warehouse, Relational model, Multidimensional model,

designing data warehouses, Model Driven Architecture.

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 2

1. Introduction

The design of data warehouses is one of the most complicated issues in business intelligence.

Its complexity is due to the proliferation of data types on the one hand, and on the other, to

the abundance of information that gave rise to the new era of Big Data. Furthermore, the data

warehouse design phase is the first and the most important step in the decision-making

process, since all other steps of the process - data transformation, data analysis, information

extraction, online analytical processing (OLAP), etc. - are largely dependent on the quality of

the designed and adopted model. For all these reasons, this issue started quickly to arouse

researchers' interest since the 1990s when the first research work appeared, which focused on

the structure of data warehouses.

Much effort has been devoted to data warehouse design, and several methods automating the

data warehouse modeling were developed but none of them has become a consensus (Gosain

and Singh, 2015). Despite the lack of a standard model, it is widely assumed that the data

warehouse design must follow the multidimensional paradigm (Kumari and Yadav, 2015) and

it must be derived from the data sources, since a data warehouse is the result of homogenizing

and integrating relevant data of the organization in a single and detailed view (Taniar and

Chen, 2011). Other research considers that user requirement analysis is crucial in data

warehouse design (Abai et al., 2013) and therefore some experts developed a new method that

supports both approaches (Battaglia et al., 2011).

In this paper, we propose a new method that follows the data-driven paradigm to design a data

warehouse from relational data sources. We opted for this approach because it permits

significant time saving since the start of the data warehouse design project requires only the

availability of the transactional data. The choice of relational data is explained by their

widespread use in all types of organizations (Ghosh, 2010).

The rest of this paper is organized as follows: in the next section, we present a set of rules

dedicated to perform the transformation from a relational data model to a data warehouse

model. Section 3 describes the transformation method by applying the set of rules previously

elaborated in the first part, before moving to describing the transformation engine. Section 4

comes as the conclusion part where we describe the perspectives of this work.

2. Related work

The design of data warehouses has been subject of several research projects. Generally,

existing approaches can be categorized into three categories: Bottom-up, top-down and mixed

approaches.

Bottom-up approaches start from a detailed analysis of data sources, but missed the decision-

makers needs. The works of (Golfarelli and Rizzi, 1998), (Moody and Kortink, 2000),

(Vrdoljak et al., 2003), (Varga, 2002) and (Sehgal and Ranga, 2016) present different

approaches that allow to generate the data warehouse schema from Entity / Association

diagrams of the data sources. The methods proposed by (Romero and Abelló, 2010) and

(Jensen et al., 2004) exploit data mining techniques and ontologies to generate the

multidimensional schema. There are even solutions suggested by big companies such as

Oracle. They proposed a set of tools that allow the transformation of logical structure to

relational structure and next transformation to Multidimensional Model of warehouse in star

or snowflake schema (Drzymala et al., 2012).

Top-down approaches (Winter and Strauch, 2003), (Annoni et al., 2006) and (Jovanovic et al.,

2014) allow building the data warehouse schema from a detailed analysis of the decision

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 3

makers needs. Verification of the correspondence between these needs and the data sources is

done a posteriori.

Mixed approaches (Phipps and Davis, 2002), (Giorgini et al., 2005) and (Abdelhedi and

Zurfluh, 2013) consider both the needs of decision makers and the source data. Therefore,

they have the advantage of designing multidimensional schemas that respect the data source

structure.

If we analyze these different methods, especially those following the data-driven approach

(Bottom-up), we can see that they are all semi-automatic, and there are even some methods

that just provide guidelines and recommendations to get suitable multidimensional schemas.

In this context, our work consists of developing a new fully automatic method called X-ETL.

This method will allow to transform a relational model into a multidimensional model without

any human intervention.

3. Rules for data warehouse design from relational data

This section presents the set of rules that we have developed from previous work (Khouri et

al., 2014) (Elmoukhi et al., 2015)(Khnaisser et al., 2015)(Santos et al., 2016) (Dahlan and

Wibowo, 2016). These rules will form the foundation of our solution to standardize the data

warehouses design.

3.1 Rules for Facts and Measures

The fact tables are the concepts of main interest for the decision making

process. They correspond to events that always occur in the organization or

company (Chandwani and Uppal, 2015) ;

The measures of the fact table should be numeric and additives (at worst

semi-additives) (Akbar et al., 2013) ;

The data of a fact table are fixed and cannot be changed (Bliujute et al.,

1998) ;

A fact table represents always a particular activity and should be

interrogated from a particular context (one or a few dimensions) ;

No line of the fact table can contain an empty value ;

A fact table contains only the foreign keys which represent the primary keys

of the dimensions and these keys must be numeric, to ensure that the fact

table is more efficient (Rudra and Nimmagadda, 2005) ;

Each combination of dimension values defines an instance of the fact table,

which is characterized by one and only one value for each measure.

Below are the mathematical representations of the rules for facts and measures:

Let TF be a fact table, MTF a fact table measures, Di a dimension of the fact table and m an

instance of MTF.

TF = P(Ev )

with:

P : Main interests

Ev : Company events ;

Let m 1 and m 2 be two instances of MTF. If m 1 and m 2 are additives:

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 4

m3 = m1 + m2

With m3 an instance of the same measure MTF ;

Suppose that f is a change function on TF

m MTF

f(m)=α

With α a constant ;

Let F be the set of fact tables and A a particular activity of the organization

For each TF F we have:

TF =A ;

TF there is at least one function f which applies at least one dimension Di

on TF ;

Let LTF be the set of rows of a fact table and l a row of LTF

l

LTF

l ∅ ;

Let fbe a foreign key and p a primary key

we have :

{f1,f2,...,fn}={p1,p2,...,pn }

k {1,2,...,n}

with fk TF and pk Di

and fk and pk of type Integer ;

Let C be the set of combinations of dimension values, c a combination of C

and f a function on MTF :

- For each instance m of M TF , the combination f( m ) C.

- For each combination c of C, the equation f(m) = c admits a unique

solution (any combination c of C admits a unique antecedent MTF) f(m)

is bijective.

3.2 Rules for Dimensions and Attributes

The dimensions determine how fact instances can be aggregated

significantly for decision making process ;

A fact table must always contain the time dimension ;

The dimensions should have numeric primary keys ;

The primary key of each dimension table should be unique (preferably auto-

increment), and fields should have an atomic value (not compound) ;

The dimension hierarchies should preferably have a simple form of 1-n

type, and avoid relationships of n-n type ;

A non-dimensional attribute contains additional information on an attribute

of the hierarchy, and is linked by to-one relationship (Golfarelli et al.,

1998) ;

The non-dimensional attributes cannot be used for aggregation (Golfarelli et

al., 1998) ;

The relationship between a fact table and a dimension is always many-to-

one(Cavalheiro and Carreira, 2016).

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 5

Below are the mathematical representations of the rules for dimensions and attributes:

Let TF be a fact table and Di a dimension of the fact table.

Let f be an aggregation function on TF

f is significant if and only if f applies one or more dimensions on the

instances of TF ;

TF (TF Dit ) with Dit a time dimension ;

Let Cp be the set of primary keys of Di

Di

Cp ∈ ℕ ;

Let Cp be the set of primary keys of Di and p1 and p2 two instances of Cp

p1 and p2

p1 p2 ;

Let R be a relationship between two dimensions

∀ R

R (n,n) ;

Let f be an aggregation function, A the set of its attributes, and an a non

dimensional attribute

For any non-dimensional attribute an we have:

an A;

Let R be a relationship between TF and Di

 TF and Di

R= (n,1).

3.3 Application of Rules

We take as an example the model below. It represents the sales activity of a set of products in

a chain of stores located in several cities and countries. The model is composed of a purchase

table related to three tables:

- The product table containing the reference, the description, the price and the type of the

product sold. We assume that a purchase concerns one and only one product, but a

product may concern several purchases;

- The customer table that contains the customers' information;

- The city table where the store is located (a city can only contain a single store).

The city table is also related to the department table to which belongs a set of cities, then the

department table to the region table and finally the region table to the country table.

Department

#Number

Name

FKRegion

Customer

#CustomerCode

LastName

FirstName

BirthDate

Gender

FKCityPC

FKCityCity

City

#PostalCode

City

FKDepartment

Purchase

#IdPurchase

Quantity

Timestamp

FKCityPC

FKCityCity

FKCustomer

FKProduct

Product

#Reference

Description

Price

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 6

Figure 1. Example of a sales transactional model

By following the rules presented in the previous sections, the fact table will be the Purchase

table. It is the table that represents a particular activity of the company and the main interest

for decision makers. It contains a numeric and additive field which can be considered as the

main measure of the fact table. In addition, this table contains the highest number of many-to-

one relationships (the highest number of foreign keys), which is the privileged type for

relations between fact tables and dimension tables (the last rule in the section 3.2).

Concerning the dimension tables, they will be respectively the Product table, the Customer

table, and the City table since they are related directly to the fact table by a one-to-many

relationship. All these tables represent the analysis contexts of the fact table and determine

how fact instances can be aggregated significantly for the decision-making process. The City

table is related to a tree of tables representing the location, and therefore they can all be

grouped together in a single table (Dim_Place). Finally, our example does not contain any

time table, so it is necessary to add one that will represent the time dimension (rule 2 in

section 3.2).

The multidimensional model below is the final result of this transformation:

Figure 2. Sales transactional model example after rules application

Thus, we chose the Purchase table as fact table, as it represents a particular business activity

and a main interest for the decision making process. This fact table contains a numeric and

additive measure, and the foreign keys of the dimensions that are also numeric. We also

added a dimension table that contains the different granularity of time, and we tried to choose

the dimensions and attributes that will enable a significant aggregation of fact instances. In

this sense, the attribute Description of the Product table was not retained as an attribute of the

dimension since it contains only a description of the product and no data that can determine

how instances of the fact table can be aggregated (cf. rule 1 for dimensions and attributes).

There is no doubt that these rules will facilitate the identification of facts, measures,

dimensions and attributes from relational data. Therefore, we have the essential components

for the construction of our own method while using the Model-driven Architecture (MDA)

techniques. Once our method will be developed, we will apply it for building a data

warehouse for the National Library of Morocco from their relational data.

Fact_Purchase

#IdPurchase

Quantity

FKPlace

FKDate

FKCustomer

FKProduct

Dim_Place

#PostalCode

Department

Region

Country

Dim_Product

#Reference

Price

Type

Dim_Date

#Timestamp

Hour

DayOfWeek

DayOfYear

Week

Month

Quarter

Semester

Year

Dim_Customer

#CustomerCode

Age

Gender

PostalCode

Department

Region

Country

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 7

4. A new method for transforming a relational model to

multidimensional model

4.1 Model Driven Architecture

MDA (Model-Driven Architecture) is a standard of the OMG (OMG, 2001), which is based

on the MDE (Model-driven engineering), providing a set of guidelines and an architecture for

the design of software systems. The MDA approach provides the opportunity to understand

complex systems and the real world through their abstraction. This abstract view of the

system is elaborated in a conceptual framework as well as a number of standards provided by

the OMG. These standards allow to define the models, their relations and their

transformations (for example: UML -Unified Modeling Language, MOF and XMI -XML

Metadata Interchange). In order to visually represent the MDA approach, the OMG has set up

a framework, structured of several types of models. Figure 3 shows the development cycle in

Y, which implements these models and their relationships:

Figure 3. The MDA process (OMG, 2001)

CIM (Computation Independent Model): he CIM allows a vision of

the system and its environment, while hiding the details of structure

and implementation. Models of the CIM level help narrow the gap

between domain experts and designers. As a result, a CIM model is

sometimes called a domain model;

PIM (Platform Independent Model): Models of the PIM level

represent a vision of system analysis and design, independently of

any technological details concerning the platform (operating system,

programming language, hardware, network performance, etc.);

PSM (Platform Specific Model): The PSM level presents a

projection of PIM level models to a specific platform. These models

combine PIM specifications with platform specific details;

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 8

PDM (Platform Description Model): These models describe the

platform on which the system will be executed, by providing a set of

technical data regarding the functionality and use of the platform.

4.2 Models transformation

Model transformation in MDA context consists to transform the PIM models to PSM models.

This process is performed by a transformation engine that applies a set of rules to PIM to

generate the PSM.

The concept of meta-model is omnipresent in this case. Thus, each model (PIM or PSM) is

based on a meta-model used to describe it. When both models use the same meta-model, it's

about an "endogenous" transformation; we talk about "exogenous" transformation in the

opposite case.

There are mainly two types of transformations:

Transformations M2M (Model to Model): used to transform models

to other models, these transformations concern all tasks to be

executed in order to get a model respecting the technical

specifications of the target environment. The standard which

technically represents this type of transformation in the MDA

approach is the MOF 2.0 QVT;

Transformations M2T (Model to Text): used to generate code or

documentation. M2T transformations constitute the MOFM2T

project which is one of many parts of MDA project.

In our case, we are interested in M2M transformation since we aim to transform a relational

model to a multidimensional model.

Basically, the main steps to complete the transformation are as follows:

1. To specify the source meta-model: first, we have to specify the

source meta-model, in our case it is the meta-model of relational

schema;

2. To specify the target meta-mo del : we must also specify the meta-

model representing the decisional concept;

3. To build the transformation engine: this step will be based on the

rules presented in section 3.

The figure below illustrates these different steps:

Conform s to

Input

Source

meta-model

Relational

meta-model

Target

meta-model

Multidimensi

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 9

Figure 4. Transformation of a relational model into a multidimensional model (Blanc and

Salvatori, 2005)

4.3 Eclipse Modeling Framework

EMF is a modeling and code generation platform that facilitates the construction of tools. It is

about a set of development tools integrated into the Eclipse environment in the form of

plugins among which we quote: the Ecore meta-model, the editor EMF.Edit, the generation

model GenModel, etc. EMF was designed to open Eclipse to the model-driven development.

It is an approach based on simplifying MOF. It allows to define meta-models then to derive

an implementation in Java to build instance models (Budinsky et al., 2009).

Figure 5 shows the architecture of the EMF Framework. The main role of this structure is to

accept models or files as input, and to generate code corresponding to tools (plug-in)

manipulating the input data.

Figure 5. EMF Architecture (Budinsky et al., 2009)

4.4 Defining the source and target meta-models

We started by defining our two meta-models (source and target) by using Ecore which is

considered as an EMF model (Eclipse Modeling Framework). In this regard, we note that all

the meta-models presented in this section are our proposal: in the ISCRAM-med conference

(El Moukhi et al., 2016), we have presented the relational metamodel and the

multidimensional metamodel that covers all multidimensional models.

The relational meta-model consists of three essential elements: a database that contains tables

which in turn contain columns. So we tried to resume these components in the figure below

(El Moukhi et al., 2016):

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 10

Figure 6. The relational meta-model

Concerning the multidimensional meta-model, it consists of a multidimensional schema that

contains facts and dimensions. In order to perform a multidimensional analysis it is necessary

to have at least two dimensions. Each fact contains measures and each dimension contains an

hierarchy of attributes. The figure 7 below resumes these components and describes our

multidimensional meta-model (target) (El Moukhi et al., 2016):

Figure 7. The multidimensional meta-model

Contrary to the previous work, this paper deals in detail with the metamodels of the various

types of multidimensional model, (Figures 8, 10, 12) and introduces a new method that allows

to transform the relational model into a multidimensional one. Thus, we have:

The star schema which contains a single fact table directly linked to

dimensions (see example of figure 2) and no dimension is related to

another, that's why we removed the reflexive link on the dimension.

Its meta-model is described below (figure 8):

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 11

Figure 8. The multidimensional meta-model for star schema

The snowflake schema which contains a single fact table with

dimensions which may be linked to other dimensions. The example

below (figure 9) illustrates this type of model.

Figure 9. The snowflake schema for sales example

The meta-model corresponding to snowflake schema is shown in the figure 10:

Fact_Purchase

#IdPurchase

Quantity

PurchasePrice

FKPlace

FKDate

FKCustomer

FKProduct

Dim_Place

#PostalCode

Department

Region

FKCountry

Dim_Product

#Reference

Name

Price

FKCategory

Dim_Date

#Timestamp

Day

FKMonth

Dim_Customer

#CustomerCode

Age

Gender

PostalCode

Department

Region

Country

Dim_Category

#IdCategory

Type

Dim_Month

#IdMonth

Month

FKYear

Dim_Country

#IdCountry

Country

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 12

Figure 10. The multidimensional meta-model for snowflake schema

The constellation schema which is the most complex. It may contain

two or many fact tables with shared dimensions, as shown in the

example below (Figure 11).

Figure 11. The constellation schema for sales example

Fact_Purchase

#IdPurchase

Quantity

PurchasePrice

FKPlace

FKDate

FKCustomer

FKProduct

Dim_Place

#PostalCode

Department

Region

Country

Dim_Product

#Reference

Name

Price

Dim_Date

#Timestamp

Hour

DayOfWeek

DayOfYear

Week

Month

Quarter

Semester

Year

Dim_Customer

#CustomerCode

Age

Gender

PostalCode

Department

Region

Country

Fact_Delivery

Amount

Volume

FKProduct

FKDate

FKSupplier

Dim_Supplier

#IdSupplier

SupplierName

Adress

FKCountry

Dim_Country

#IdCountry

CountryName

CodeName

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 13

Figure 12. The multidimensional meta-model for constellation schema

In order to verify the conformity of our models to these meta-models, we created two other

files with xsd language in which we described how the model structure should be to comply

with its meta-model. Thus, we got an xsd file to validate the relational model and another one

to validate the multidimensional model. The codes fragments below correspond to these two

files.

The XML Schema Definition for relational models :

!"#$%&'()*+,- "1.0" "./

!#)0)12'$3#$%,)0#)- "http://www.w3.org/2001/XMLSchema"344(*564'7+($8'936%4- "unquali

fied"'%'$',47+($8'936%4- "qualified"./

!#)0'%'$',4,3$'- "relationalSchema"4:;'- "relationalSchemaType" <./

!#)01+$;%'#=:;',3$'- "columnType" ./

!#)0)*$;%'>+,4',4 ./

!#)0'#4',)*+,53)'- "xs:string" ./

!#)0344(*564'4:;'- "xs:string",3$'- "name"6)'- "optional" <./

!#)0344(*564'4:;'- "xs:string",3$'- "isPk"6)'- "optional" <./

!#)0344(*564'4:;'- "xs:string",3$'- "type"6)'- "optional" <./

!#)0344(*564'4:;'- "xs:string",3$'- "isFk"6)'- "optional" <./

!<#)0'#4',)*+, ./

!<#)0)*$;%'>+,4',4 ./

!<#)01+$;%'#=:;' ./

!#)01+$;%'#=:;',3$'- "columnsType" ./

!#)0)'?6',1' ./

!#)0'%'$',44:;'- "columnType",3$'- "column"$3#@116()- "unbounded"$*,@116()- "0" <./

!<#)0)'?6',1' ./

!<#)01+$;%'#=:;' ./

!#)01+$;%'#=:;',3$'- "associationType" ./

!#)0)*$;%'>+,4',4 ./

!#)0'#4',)*+,53)'- "xs:string" ./

!#)0344(*564'4:;'- "xs:string",3$'- "multiplicity"6)'- "optional" <./

!#)0344(*564'4:;'- "xs:string",3$'- "target"6)'- "optional" <./

!<#)0'#4',)*+, ./

!<#)0)*$;%'>+,4',4 ./

!<#)01+$;%'#=:;' ./

!#)01+$;%'#=:;',3$'- "associationsType" ./

!#)0)'?6',1' ./

!#)0'%'$',44:;'- "associationType",3$'- "association"$3#@116()- "unbounded"$*,@116()-

"0"<./

!<#)0)'?6',1' ./

!<#)01+$;%'#=:;' ./

!#)01+$;%'#=:;',3$'- "tableType" ./

!#)0)'?6',1' ./

/

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 14

The XML Schema Definition for multidimensional models

!"#$%&'()*+,- "1.0" "./

!#)0)12'$3#$%,)0#)- "http://www.w3.org/2001/XMLSchema"344(*564'7+($8'936%4- "unquali

fied"'%'$',47+($8'936%4- "qualified"./

!#)0'%'$',4,3$'- "multidimensionalSchema"4:;'- "multidimensionalSchemaType" <./

!#)01+$;%'#=:;',3$'- "measureType" ./

!#)0)*$;%'>+,4',4 ./

!#)0'#4',)*+,53)'- "xs:string" ./

!#)0344(*564'4:;'- "xs:string",3$'- "name" <./

!<#)0'#4',)*+, ./

!<#)0)*$;%'>+,4',4 ./

!<#)01+$;%'#=:;' ./

!#)01+$;%'#=:;',3$'- "foreignkeyType" ./

!#)0)*$;%'>+,4',4 ./

!#)0'#4',)*+,53)'- "xs:string" ./

!#)0344(*564'4:;'- "xs:string",3$'- "name"6)'- "optional" <./

!<#)0'#4',)*+, ./

!<#)0)*$;%'>+,4',4 . /

!<#)01+$;%'#=:;' ./

!#)01+$;%'#=:;',3$'- "fieldsType" ./

!#)0)'?6',1' ./

!#)0'%'$',44:;'- "measureType",3$'- "measure" <./

!#)0'%'$',44:;'- "foreignkeyType",3$'- "foreignkey"$3#@116()- "unbounded"$*,@116()- "0

"<./

!<#)0)'?6',1' ./

!<#)01+$;%'#=:;' ./

!#)01+$;%'#=:;',3$'- "associationType" ./

!#)0)*$;%'>+,4',4 ./

!#)0'#4',)*+,53)'- "xs:string" ./

!#)0344(*564'4:;'- "xs:string",3$'- "multiplicity"6)'- "optional" <./

!#)0344(*564'4:;'- "xs:string",3$'- "target"6)'- "optional" <./

!<#)0'#4',)*+, ./

!<#)0)*$;%'>+,4',4 ./

!<#)01+$;%'#=:;' ./

!#)01+$;%'#=:;',3$'- "associationsType" ./

!#)0)'?6',1' ./

!#)0'%'$',44:;'- "associationType",3$'- "association"$3#@116()- "unbounded"$*,@116()-

"0"<./

!<#)0)'?6',1' ./

!<#)01+$;%'#=:;' ./

!#)01+$;%'#=:;',3$'- "factType" ./

!#)0)'?6',1' ./

!#)0'%'$',44:;'- "fieldsType",3$'- "fields" <./

!#)0'%'$',44:;'- "associationsType",3$'- "associations" <./

!<#)0)'?6',1' ./

!#)0344(*564'4:;'- "xs:string",3$'- "name" <./

!<#)01+$;%'#=:;' ./

!#)01+$;%'#=:;',3$'- "attributeType" ./

!#)0)*$;%'>+,4',4 ./

!#)0'#4',)*+,53)'- "xs:string" ./

!#)0344(*564'4:;'- "xs:string",3$'- "name"6)'- "optional" <./

!#)0344(*564'4:;'- "xs:string",3$'- "isPk"6)'- "optional" <./

!#)0344(*564'4:;'- "xs:string",3$'- "type"6)'- "optional" <./

!#)0344(*564'4:;'- "xs:string",3$'- "isFk"6)'- "optional" <./

Many other meta-models have been proposed in the literature. We quote as examples

(Hachaichi and Feki, 2013) (Srai et al., 2017) (Sapia et al., 1999) (Choura and Feki, 2011).

Concerning the relational meta-model, it is a well-known standard that is repeated in almost

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 15

all research with some slight differences (Lämmel, 2005) (Chang et al., 2003) (Inria, 2005).

As for the multidimensional meta-model, we find that each existing research work presents a

new schema but the common point between all these works is that these meta-models remain

too general and do not treat each type of multidimensional model separately. We take as an

example the CWM schema (Figure 13) which represents the most known and the most used

multidimensional meta-model in the field.

Figure 13. Multidimensional Metamodel of CWM (OMG, 2003)

As we can see, it is a meta-model that covers all types of multidimensional models and

therefore there will be fewer restrictions when we choose to work with a specific type. In this

context, our work comes to complete these research works by treating each type of

multidimensional model (star-flocon-constellation) separately and presenting its meta-model.

Thus, we obtained for the snowflake schema (figure 10) a cardinality of three for the

dimension table, since we must have at least one more dimension related to another. For the

constellation schema (figure 12), we must have at least two fact tables related by dimensions,

which explains the cardinality of two for the fact table.

4.5 The transformation engine

After defining our meta-models, we built the transformation engine based on the rules

presented in section 2 and principles of the MDA. Firstly, we developed a Java program that

calculates the number of foreign keys in each table of the relational model, which allowed us

to detect the (potential) fact tables.

Once the fact tables are identified, we generate a multidimensional model that contains only

the fact table with the dimension tables that are directly related to it in the relational model.

After that, we will also include the indirectly related dimensions in a future work.

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 16

The figure 14 shows the architecture of the X-ETL project.

Figure 14. X-ETLp roject architecture

The most important part of the project is the file that allows parsing tables in order to

calculate the number of foreign keys and then identify potential fact tables. The figure 15

shows the edit window for this file.

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 17

Figure 15. Dom parsing of table element

Once the fact tables are detected, the second step consists of detecting the dimensions by

parsing associations of the relational model to identify many-to-one relationships. The figure

16 represents the Dom parsing of association element.

Figure 16. Dom parsing of association element

The complex calculations that are made on the cardinalities to detect the dimensions justify

the choice of the JAVA language, since the other transformation languages such as ATL

(Atlas Transformation Language) or QVT (Query/View/Transformation) do not allow tomake

this kind of complicated calculation.

After detecting the dimension tables, the file described in Figure 17 allows to create them in

the target multidimensional model.

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 18

Figure 17. Creating a new dimension

The diagram below illustrates all the steps of the X-ETL transformation.

Figure 18. The steps of the X-ETL transformation

Below is a screenshot of the X-ETL application.

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 19

Figure 19. Screenshot of the X-ETL engine

5. Conclusion

In this paper, we presented a new data-driven method for designing a multidimensional model

from a relational model. This method is mainly based on a list of rules to identify the different

elements of the multidimensional schema and consists of two steps. The first one aims to

identify fact tables by calculating the number of foreign keys in each table of the relational

model, and the second one allows identifying dimensions that are directly related to the fact

table, by analyzing the cardinalities of relations. At the end, several multidimensional models

are generated in an automatic way. At this stage, it should be noted that the quality of these

generated models depends greatly on the quality of the source model, and therefore it is very

important to verify the relational source model before using the X-ETL engine. Our future

work will consist of finding a method to select dimensions from tables that are indirectly

related to the fact table in order to generate a complete multidimensional model.

References

Abai, N.H.Z., Yahaya, J.H., Deraman, A., (2013). User Requirement Analysis in Data

Warehouse Design: A Review. Procedia Technology 11, pp. 801–806.

doi:10.1016/j.protcy.2013.12.261

Abdelhédi, F., Zurfluh, G., (2013). User support system for designing decisional database, in :

Proceedings of the 6th International Conference on Advances in Computer-Human

Interactions. Nice, France, pp. 377–382.

Akbar, K., Krishna, S.M., Reddy, T.V.S., (2013). ETL process modeling in DWH using

enhanced quality techniques. International Journal of Database Theory & Application 6,

pp. 179–197.

Annoni, E., Ravat, F., Teste, O., Zurfluh, G., (2006). Towards multidimensional requirement

design, in: Proceedings of the 8th International Conference on Data Warehousing and

Knowledge Discovery. Springer-Verlag, Krakow, Poland, pp. 75–84.

Battaglia, A., Golfarelli, M., Rizzi, S., (2011). Qbx: a case tool for data mart design, in:

International Conference on Conceptual Modeling. Springer, pp. 358–363.

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 20

Blanc, X., Salvatori, O., (2005). MDA en action: ingénierie logicielle guidée par les modèles.

Eyrolles.

Bliujute, R., Saltenis, S., Slivinskas, G., Jensen, G., (1998). Systematic change management

in dimensional data warehousing. Citeseer.

Budinsky, F., Merks, E., Paternostro, M., Steinberg, D., (2009). EMF: Eclipse Modeling

Framework. Addison-Wesley,

Cavalheiro, J., Carreira, P., (2016). A multidimensional data model design for building energy

management. Advanced Engineering Informatics, Vol. 30, issue 4, pp. 619–632. doi:

10.1016/j.aei.2016.08.001

Chandwani, G., Uppal, V., (2015). Implementation of Star Schemas from ER Model.

International Journal of Database Theory and Application 8, pp. 111–130.

doi:10.14257/ijdta.2015.8.3.10

Chang, D., Mellor, D., Poole, J., Tolbert, D., (2003). Common Warehouse Metamodel:

Developer's Guide. Wiley, p. 86.

Choura, H., Feki, J., (2011). MDA Compliant Approach for Data Mart Schemas Generation,

in: Proceedings of the First international conference on Model and data engineering.

Springer, pp. 262–269.

Dahlan, A., Wibowo, F. W., (2016). Design of Library Data Warehouse Using SnowFlake

Scheme Method, in: 7th International Conference on Intelligent Systems, Modelling and

Simulation (ISMS). IEEE, pp. 318–322. doi: 10.1109/ISMS.2016.71

ElMoukhi, N., El Azami, I., Mouludi, A., (2016). X-ETL Engine: from relational model to a

multidimensional model, in: 3rd International Conference on Information Systems for

Crisis Response and Management in Mediterranean Countries (ISCRAM). Madrid,

Spain, pp. 39–42.

El Moukhi, N., El Azami, I., Mouludi, A., (2015). Data warehouse state of the art and future

challenges, in: International Conference on Cloud Technologies and Applications

(CloudTech). IEEE. Doi : 10.1109/CloudTech.2015.7337004.

Ghosh, D., (2010). Multiparadigm data storage for enterprise applications. IEEE software 27,

pp. 57–60.

Giorgini, P., Rizzi, S., Garzetti, M., (2005). Goal-oriented requirement analysis for data

warehouse design, in: Proceedings of the 8th ACM International Workshop on Data

Warehousing and OLAP. ACM, Bremen, Germany, pp. 47–56.

Golfarelli, M., Maio, D., Rizzi, S., (1998). Conceptual design of data warehouses from E/R

schemes, in: Proceedings of the Thirty-First Hawaii International Conference on System

Sciences. IEEE, pp. 334–343.

Gosain, A., Singh, J., (2015). Conceptual Multidimensional Modeling for Data Warehouses:

A Survey, in:Proceedings of the 3rd International Conference on Frontiers of Intelligent

Computing: Theory and Applications (FICTA). Springer International Publishing, pp.

305–316.

Hachaichi, Y., Feki, J., (2013). An automatic method for the design of multidimensional

schemas from object oriented databases. International Journal of Information Technology

& Decision Making, Vol. 12, issue 6, pp. 1223–1259.

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 21

Inria, (2005). ATL transformation example: class to relational.

https://www.eclipse.org/atl/atlTransformations/Class2Relational/ExampleClass2Relation

al[v00.01].pdf

Jensen, M.R., Holmgren, T., Pedersen, T.B., (2004). Discovering Multidimensional Structure

in Relational Data, in: Proceedings of the 6th International Conference on Data

Warehousing and Knowledge Discovery. Springer Berlin Heidelberg, Berlin, Heidelberg,

pp. 138–148.

Jovanovic, P., Romero, O., Simitsis, A., Abelló, A., Mayorova, D., (2014). A requirement-

driven approach to the design and evolution of data warehouses. Information Systems,

Vol. 44, pp. 94–119. https://doi.org/10.1016/j.is.2014.01.004

Khnaisser, C., Lavoie, L., Diab, H., Ethier, J.-F., (2015). Data Warehouse Design Methods

Review: Trends, Challenges and Future Directions for the Healthcare Domain, in : East

European Conference on Advances in Databases and Information Systems (ADBIS).

Springer, pp. 76–87. doi: 10.1007/978-3-319-23201-0_10

Khouri, S., Bellatreche, L., Jean, S., Ait-Ameur, Y., (2014). Requirements Driven Data

Warehouse Design: We Can Go Further, in: International Symposium On Leveraging

Applications of Formal Methods, Verification and Validation (ISoLA). Springer, pp. 588–

603.

Kumari, S., Yadav, P., (2015). Study of Influence of Data Mining & Data Warehousing, in:

Proceedings of National Conference on Innovative Trends in Computer Science

Engineering (ITCSE). IJRRA, pp. 138–140.

Lämmel, R., Saraiva, J., Visser, J., (2005). Generative and Transformational Techniques in

Software Engineering. Springer, p. 41.

Moody, D.L., Kortink, M.A.R., (2000). From enterprise models to dimensional models: a

methodology for data warehouse and data mart design., in: Proceedings of the

International Workshop on Design andManagement of Data Warehouses.St ockholm,

Sweden, p. 5.

OMG, (2003). Common Warehouse Metamodel (CWM) Specification. Vol. 1, Version 1.1,

pp. 8-3.

OMG, (2001). MDA Specifications. https://www.omg.org/mda/specs.htm

Phipps, C., Davis, K.C., (2002). Automating data warehouse conceptual schema design and

evaluation., in: Proceedings of the 4th International Conference on Design and

Management of Data Warehouses. Toronto, Canada, pp. 23–32.

Romero, O., Abelló, A., (2010). A framework for multidimensional design of data

warehouses from ontologies. Data & Knowledge Engineering, Vol. 69, pp. 1138–1157.

https://doi.org/10.1016/j.datak.2010.07.007

Rudra, A., Nimmagadda, S.L., (2005). Roles of multidimensionality and granularity in

warehousing Australian resources data, in: Proceedings of the 38th Annual Hawaii

International Conference on System Sciences. IEEE, pp. 216b–216b.

Santos, M. Y., Oliveira e Sá, J., (2016). A Data Warehouse Model for Business Processes

Data Analytics, in: International Conference on Computational Science and Its

Applications (ICCSA). Springer, pp. 241–256. doi: 10.1007/978-3-319-42092-9_19

Towards a new automatic data warehouse design method, Nawfal El Moukhi, Ikram El Azami, Abdelaaziz Mouloudi &Abdelali Elmounadi

e-TI, Revue électronique des Technologies de l'Information. http://www.revue- eti.net, Numéro 11. 2018. ISSN 1114-8802 22

Sapia, C., Blaschka, M., Höfling, G., Dinter, B., (2017). Extending the E/R Model for the

Multidimensional Paradigm, in: Proceedings of the Workshops on Data Warehousing

and Data Mining: Advances in Database Technologies. Springer, pp. 105–116.

Sehgal, S., Ranga, K. K., (2016). Translation of Entity Relational Model to Dimensional

Model. International Journal of Computer Science and Mobile Computing, Vol. 5, issue

5, pp. 439–447.

Srai, A., Guerouate, F., Berbiche, N., Drissi, H., (2017). An MDA approach for the

development of data warehouses from Relational Databases Using ATL Transformation

Language. International Journal of Applied Engineering Research, Vol. 12, issue 12, pp.

3532–3538.

Taniar, D., Chen, L. (Eds.), (2011). Integrations of Data Warehousing, Data Mining and

Database Technologies: Innovative Approaches. IGI Global.

Varga, M., (2002). A Procedure of Conversion of Relational into Multidimensional Database

Schema. Journal of Computing and Information Technology, Vol. 10, pp. 69–84.

Vrdoljak, B., Banek, M., Rizzi, S., (2003). Designing web warehouses from XML schemas,

in: International Conference on Data Warehousing and Knowledge Discovery. Springer,

pp. 89–98.

Wiak, S., Drzymala, P., Welfle, H., (2012). Using ORACLE tools to generate

Multidimensional Model in Warehouse. Przegląd Elektrotechniczny, pp. 257–262.

Winter, R., Strauch, B., (2003). A method for demand-driven information requirements

analysis in data warehousing projects, in: Proceedings of the 36th Annual Hawaii

International Conference on System Sciences. IEEE.

A data warehouse (DW) is a vast repository of data that facilitates decision-making for businesses and companies. This concept dates back to the 1980s and it has been widely accepted. One of the key points for the success of the process of data warehousing lies in the definition of the warehouse model depending on data sources and analysis needs. Once the data warehouse is designed, the content and structure of the data sources, as well as the requirements analysis are required to evolve, therefore, an evolution of the model must take place (diagram and data). In this context, several approaches have been developed to design and implement data warehouses. Nevertheless, there is no standard process that deals with designing all of the data warehouse layers, also, there is no software that encompasses this type of problem. In general, the majority of these approaches focus on a particular aspect of data warehouse such as data storage, ETL process, OLAP, reporting, etc, and does not cover its entire lifecycle. A Model-Driven Architecture (MDA) is a standard approach, its aims to support all phases of software manufacturing by promoting the use of models and the transformations between them. Moreover, this approach aims to automate the process of software engineering, thereby decreasing the cost of software development and enhancing its productivity. In this study, we present a systematic review of various works on the data warehouse design methods. We compare and discuss these works according to the criteria that seem relevant for this issue. We present a new design approach for multidimensional schemas construction from relational models using MDA techniques, we also develop the resulting research perspectives.

The existence of the library as a technical service unit in a campus is very important to provide services to the academic community. With increasing number of library collections, database that is built has to be able to improve the services oriented on providing data warehouse. Especially at managerial level requires complete information, quickly and accurately to support the process and planning, evaluation, and right decision-making. The design of data warehouse is determined by description of proper information requirement, selection of valid data source, design of data warehouse and ETL process to integrate, extract, cleanse, transform and populate it into data warehouse. The snowflake scheme design method has been applied to accommodate dimension tables of database and other dimension sub-table, so it can generate more information that would be used as material to make a decision.

These last years, the sources of data became very heterogeneous and massive. The term "big data" is born to indicate this phenomenon. Unfortunately, he becomes difficult to manage big data to power the decisional system, and also it increases the time of interrogation of data from data warehouses. Several research works focused recently, on the proposal for the architectures of data warehouses for this type of data ''big data'', and on the implementation of new algorithms of interrogation of these warehouses to improve the time of the answers. This article proposes a Model-Driven Architecture (MDA) approach for the development of data warehouses independently of any execution platform, to allow the facilitation of the development of these data warehouses as well as the migration of information systems based on relational DBMS to systems NoSQL.

Data organization is a critical aspect in Building Energy Data Management. Yet, despite the importance of the topic, no sound reference model for energy data has been proposed in the literature that has been developed according to well-founded methodologies. This article proposes a reference data model developed according to standard multidimensional modeling methodologies and improved iteratively in review meetings with expert users (in the building energy management domain). The quality of the model is evaluated according to complexity, usability, and design metrics thus achieving a high-quality re-usable multidimensional data model that can be applied to create or improve on the data model designs of building energy management systems.

Business Process Management and Business Intelligence initiatives are commonly seen as separated organizational projects, suffering from lack of coordination, leading to a poor alignment between strategic management and operational business processes execution. Researchers and professionals of information systems have recognized that business processes are the key for identifying the user needs for developing the software that supports those needs. In this case, a process-driven approach could be used to obtain a Data Warehouse model for the Business Intelligence supporting software. This paper presents a process-based approach for identifying an analytical data model using as input a set of interrelated business processes, modeled with Business Process Model and Notation version 2.0, and the corresponding operational data model. The proposed approach ensures the identification of an analytical data model for a Data Warehouse repository, integrating dimensions, facts, relationships and measures, providing useful data analytics perspectives of the data under analysis.

The paper presents modern techniques of data modelling and processing, collected by the company. It presents the process of multidimensional data modelling (include the transformation of logical structure to relational structure and next transformation to Multidimensional Model of warehouse in star or snowflake schema). It also shows the ETL process and methods of creating OLAP cubes by use of ORACLE tools to support decision making by business analysts. An approach based on data mining techniques allows analysts to capture certain features in customers, to offer dedicated products for the customer groups. Based on customer behaviour can be concluded about his tendencies to their certain behaviours and preferences.

In secondary data use context, traditional data warehouse design methods don't address many of today's challenges; particularly in the healthcare domain were semantics plays an essential role to achieve an effective and implementable heterogeneous data integration while satisfying core requirements. Forty papers were selected based on seven core requirements: data integrity, sound temporal schema design, query expressiveness, heterogeneous data integration, knowledge/source evolution integration, traceability and guided automation. Proposed methods were compared based on twenty-two comparison criteria. Analysis of the results shows important trends and challenges, among them (1) a growing number of methods unify knowledge with source structure to obtain a well-defined data warehouse schema built on semantic integration; (2) none of the published methods cover all the core requirements as a whole and (3) their potential in real world is not demonstrated yet.

Data warehouses (\(\mathcal{D}\mathcal{W}\)) are defined as data integration systems constructed from a set of heterogeneous sources and user's requirements. Heterogeneity is due to syntactic and semantic conflicts occurring between used concepts. Existing \(\mathcal{D}\mathcal{W}\) design methods associate heterogeneity only to data sources. We claim in this paper that heterogeneity is also associated to users' requirements. Actually, requirements are collected from heterogeneous target users, which can cause semantic conflicts between concepts expressed. Besides, requirements can be analyzed by heterogeneous designers having different design skills, which can cause formalism heterogeneity. Integration is the process that manages heterogeneity in \(\mathcal{D}\mathcal{W}\) design. Ontologies are recognized as the key solution for ensuring an automatic integration process. We propose to extend the use of ontologies to resolve conflicts between requirements. A pivot model is proposed for integrating requirements schemas expressed in different formalisms. A \(\mathcal{D}\mathcal{W}\) design method is proposed for providing the target \(\mathcal{D}\mathcal{W}\) schema (star or snowflake schema) that meets a uniformed and consistent set of requirements.