Towards Data Warehouse Schema Design from Social Networks
Dynamic Discovery of Multidimensional Concepts
Rania Yangui
1
, Ahlem Nabli
2
and Faiez Gargouri
1
1
Institute of Computer Science and Multimedia, Sfax University, BP 1030, Sfax, Tunisia
2
Faculty of Sciences, Sfax University, BP 1171, Sfax, Tunisia
Keywords:
Data Warehouse, Social Network, Ontology, Flexibility, Scalability, Dynamicity, Clustering.
Abstract:
This research work is conducting as part of the project BWEC (Business for Women in Women of Emerging
Country) that aims to improve the socio-economic situation of handicraft women by providing true techno-
logical means. In fact, since few years, the Web has been transformed into an exchange platform where users
have become the main suppliers of information through social media. User-generated data are usually rich
and thus need to be analyzed to enhance decision. The storage and the centralization of these data in a data
warehouse (DW) are highly required. Nevertheless, the growing complexity and volumes of the data to be
analyzed impose new requirements on DW.
In order to address these issues, in this paper, we propose four stages methodology to define a DW schema
from social networks. Firstly we design the initial DW schema based on the existing approaches. Secondly,
we apply a set of transformation rules to prepare the creation of the NOSQL(Not Only SQL) data warehouse.
Then, based on user’s requirement, clustering of social networks profiling data will be performed which allows
the dynamic discovery of multidimensional concepts. Finally, the enrichment of the NoSQL DW schema by
the discovered MC will be realized to ensure the DW schema evolution.
1 INTRODUCTION
Because of the rapid development of the Internet,
the availability of various types of data has increased
tremendously. In fact, the creation of many sites (such
as Facebook, LinkedIn, Twitter, etc.) and forums
has made the users perceive the Web as a place in
which they exchange ideas, opinions as well as con-
tents. However, if these tools make the sharing and
collaboration between the users easy, they may cause
new challenges concerning the relevant exploitation
of the produced data. Analyzing, understanding, as
well as managing the huge volumes of complex data
produced from the social networks (SN) broach a
paramount importance and draw the attention of many
researchers. In fact, the companies expect to acquire
important information from this data so as to improve
their marketing. This is the case of handicraft women
in the BWEC
1
project. This latter aims at improving
the socio-economic situation of these women by pro-
viding true technological means in concordance with
1
Towards a new Manner to use Affordable Technologies
and Social Networks to Improve Business for Women in
Emerging Countries (http://projetat.cerist.dz/)
the women habits and the technical context of their
countries. The stages proposed to accomplish the pur-
pose of this project are summarized in (Figure 1).
The use of online SN can play a very important
role in the social and economic development of this
population. For instance, SN can be used not only
to promote their products but also to enhance the
brand value. It can also be used to strengthen con-
sumer relations as well as to improve the quality of
the services and products through receiving feedback
from the market itself. Thus, the establishment of a
decision-making process has proved to be necessary.
Originally developed for the needs of support de-
cision, data warehouses (DW) have proven to be an
adequate solution to a variety of applications and
fields. DW contains all the information integrated
from heterogeneous sources into multidimensional
schema to enhance data access for both analysis and
decision making.
Many methodologies can be used to create a
DW (Nabli, 2013) which are demand-driven method-
ologies, data-driven methodologies and Mixed (de-
mand/data driven) methodologies. These warehous-
ing methodologies have shown their efficiency when
338
Yangui R., Nabli A. and Gargouri F..
Towards Data Warehouse Schema Design from Social Networks - Dynamic Discovery of Multidimensional Concepts.
DOI: 10.5220/0005383903380345
In Proceedings of the 17th International Conference on Enterprise Information Systems (ICEIS-2015), pages 338-345
ISBN: 978-989-758-096-3
Copyright
c
2015 SCITEPRESS (Science and Technology Publications, Lda.)
Figure 1: Project Steps.
dealing with structured data. Its require measures
and dimensions of a DW to be known at the design
stage (N. U. Rehman and Scholl, 2012). However,
the growing complexity and volumes of the data to be
analyzed impose new requirements on data warehous-
ing systems.
As argued by many works (N. U. Rehman and
Scholl, 2012) (E. Gallinucci and Rizzi, 2013), exist-
ing warehousing methodologies cannot be success-
fully applied to handle the above-mentioned chal-
lenges. So, a remarkable effort must be made to inte-
grate the huge amount of complex data from SN and
make them accessible to OLAP (On-Line Analytical
Processing) and analyzing tools.
Our challenges can be summarized by the follow-
ing questions: how to determine multidimensional
concepts (MC) from social data? How to adjust the
DW to take into account these new concepts? How to
overcome the problems caused by the heterogeneity
of the data? How to take into account the dynamicity
of user-generated data as well as the users needs in an
existing DW?
This paper tends to answer the mentioned ques-
tions by proposing a methodology to design a DW
schema via content-based dynamic discovery of MC
from social networks generated data. The proposed
methodology will be applied on real case study. This
case study looks at handicraft women social networks.
Our contributions focus on the DW schema evolv-
ing to manage data about handicraft women from SN.
In this context, DW schema modeling is a complex
task which involves knowledge of SN structure and
familiarity with DW technologies. What makes this
task even more challenging is the fact that social data
continue to grow rapidly and analytical requirements
change over time. Given these challenges, the semi-
automatic modeling of the warehouse schema is re-
quired.
This paper is organized as follows. Section 2 re-
views some related works concerning the DW schema
creation from SN and the semi-automatic generation
of DW schema. Section 3 overviews our method-
ology. Sections 4 describes the initial DW schema
creation. Sections 5 details the dynamic discovery of
MC. Section 6 concludes the paper and draws future
research directions.
2 RELATED WORKS
Data generated from SN are usually rich and need to
be analyzed to support decision. The storage and the
centralization of these data in a DW are highly re-
quired. However as mentioned above, warehousing
methods have shown some limitations when dealing
with SN data. Challenged by these limitations, DW
researchers put tremendous efforts into extending its.
2.1 Multidimensional Modeling from
Social Networks
In the literature, many researchers have proposed ap-
proaches for semi-automatic modeling of DW.
For example, a data warehousing architecture for
analyzing large data sets at Facebook, used for friend
recommendation, is described by (A. Thusoo and Liu,
2010). The authors describe the challenges of im-
plementing a DW for data intensive Facebook appli-
cations and present a number of contributed open-
source technologies for warehousing petabytes of
data. These include Scribe
2
, Hadoop
3
and Hive
4
which together form the cornerstones of the log col-
lection, storage and analytic infrastructure at Face-
book.
The paper mainly focuses on flexibility and scal-
ability issues. However no insight on the underlying
models is given and no dynamic discovery of MC is
mentioned.
(P. Kazienko and Brdka, 2011) focus on develop-
ing a conceptual generic model for multidimensional
SN that allows capturing information about different
types of activities and interactions between users. It
also represents the dynamics of users behavior. The
proposed model encompasses information about the
different relations and groups that exist within a given
relation layer and in a specific time window. The pro-
posed model covers three main dimensions: relation
layers, time windows and groups. Social groups are
extracted by means of clustering methods. However,
2
http://wiki.github.com/facebook/scribe
3
http://wiki.apache.org/hadoop
4
http://wiki.apache.org/hadoop/Hive
TowardsDataWarehouseSchemaDesignfromSocialNetworks-DynamicDiscoveryofMultidimensionalConcepts
339
the proposed model doesnt manage the flexibility and
the scalability of social networks data.
(N. U. Rehman and Scholl, 2012) provide a DW
solution for hosting the public data stream of Twit-
ter messaging. The authors enrich the multidimen-
sional analysis of such data via content-driven discov-
ery of dimensions and classifying hierarchies. In the
first step, data mining algorithms are applied to clus-
ter dimensional data. In the second step, the acquired
classification is added as a new aggregation path to
the respective dimension, leading to the third step of
enabling this new aggregation path in OLAP queries.
Nevertheless, this work is limited to the granularity
level addition and ignores the other MC such as facts
and dimensions. Moreover, the proposed model is in-
flexible and no scalable.
(E. Gallinucci and Rizzi, 2013) propose a method-
ology called meta-stars to model topic hierarchies
in ROLAP systems. Its basic idea is to use meta-
modeling coupled with navigation table and with tra-
ditional dimension tables. The navigation tables sup-
port hierarchy instances with different lengths and
with non-leaf facts, and allow different roll-up seman-
tics to be explicitly annotated. The meta-modeling
enables hierarchy heterogeneity and dynamics to be
accommodated. However, this work is based on a re-
lational approach which presents limitations regard-
ing to schema scalability.
(Moalla and Nabli, 2014) present a method to
multidimensional schema construction from unstruc-
tured data extracted from SN. This construction is car-
ried out from Facebook page in order to analyze the
customers opinions. A real case study has been devel-
oped to illustrate the proposed method and to confirm
that the SN analysis can predict the success prospects
of the products. Nevertheless, the dynamic discovery
of MC is not supported. The proposed model is not
flexible and not adaptable to the huge amount of so-
cial data.
Based on the previous study, most of the works
show no indication of the dynamic determination of
MC seen the velocity of SN data. Also the DW
schemas are generally fixed at design stage.
2.2 Dynamic Discovery of
Multidimensional Concepts
Nowadays, we are experiencing a rapid growth of so-
cial structures supported by communication technolo-
gies and various Web-based services. Due to scale,
complexity and dynamicity, user-generated data from
SN are very difficult to store and analyze in terms of
traditional data warehousing methods (N. U. Rehman
and Scholl, 2012). To overcome these problems,
many authors have worked on dynamic discovery of
MC and have used data mining to build a DW.
In this context, (Usman and Pears, 2011) provide
a methodology to design semi-automatically DWs
schema with hierarchical clustering. This latter is
used to perform a pre-processing on the data. After
that, the system identifies both facts and dimensions
into the clustered data.
Rehman proposes a system to dynamically build
hierarchies based on data from Twitter (N. U. Rehman
and Scholl, 2012). This paper has two Interests: a)
The cube is built on original data which are the mes-
sages of users on a SN. b) Data mining is used to
dynamically build hierarchies. Thanks to data min-
ing, the categories of network users described in hi-
erarchies are updated automatically. On the other
hand, Ceci uses a hierarchical clustering to integrate
continuous variables as dimensions in a DW schema
(M. Ceci and Malerba, 2011). It discretizes a continu-
ous dimension so that the user can perform operations
on existing querying a cube: Roll-up and DrillDown.
As for the current work, (L. Sautot and Molin,
2014) propose using hierarchical agglomerative clus-
tering with a metric that comes from ecological stud-
ies to build semi-automatically hierarchical dimen-
sions in an OLAP cube. The authors perform a hierar-
chical clustering on heterogeneous data sets that con-
tains qualitative and quantitative variables. They offer
a prototypical automatic system which builds dimen-
sion for an OLAP cube and measure the performances
of this system according to the number of clustered
individuals and according to the number of variables
used for clustering.
Table 1 highlights a summary of the literature re-
view which is based on seven criteria (Concept M.:
Conceptual Model, D. MC: multidimensional con-
cepts, Methodology, SN: Social Network, Ontology,
Flexibility, Scalability).
All the mentioned works present several interest-
ing mining. It has been recognized that mining tech-
niques such as Clustering can help in designing DW
schema. That is why we adopt this orientation for
the dynamic discovery of MC. However, no work has
ever dealt with the semantic heterogeneity. More-
over, no work has ever followed a mixed approach
(data/demand driven approach). Furthermore, it is
worth noting that just one work has provided the scal-
ability of the schema. At the same time, the hetero-
geneity and the growth of the social data need to be
considered in order to properly retrieve needed data.
The frequent arrival of new needs requires that the
system should be adaptable to changes.
Based on the above discussion, there is a strong
need of a significant methodology that allows a dy-
ICEIS2015-17thInternationalConferenceonEnterpriseInformationSystems
340
Table 1: Summary of the literature review.
Work Concept M. D. MC Methodology SN Ontology Flexibility Scalability
(Thusoo, 2010) - No Data driven Yes No Yes Yes
(Ceci, 2011) - Yes Demand driven No No No No
(Usman, 2011) - Yes Data driven No No No No
(Kazienko, 2011) - Yes Data driven Yes No No No
(Rehman, 2012) x-DFM Yes Data driven Yes No No No
(Gallinucci, 2013) Meta-Star Yes Data driven Yes No No No
(Moalla, 2014) Star Schema No Mixed Yes No No No
(Sautot, 2014) Constellation Yes Data driven No No No No
Our Proposition x-DFM Yes Mixed Yes Yes Yes Yes
namic discovery of MC based on ontology. The mod-
eling of a flexible and scalable DW schema is also
required to deal with data from SN.
3 OVERVIEW OF THE
PROPOSED METHODOLOGY
In this paper, we propose four stages to define data
warehouse methodology from social networks (Fig-
ure 2). We will begin with the creation of the ini-
tial DW schema from structured and heterogeneous
sources following a classical approach. The second
stage will relate to the transformation of the DW
schema into a NoSQL Data Base. After that, we will
dynamically determine the MC, their types as well
as their locations. These MC are used to enrich the
NoSQL DW schema.
Our methodology takes advantages of the matu-
rity of existing design approaches, the scalability of
NOSQL Data Base and the capability of the dynamic
discovery of multidimensional concepts through clus-
tering techniques. Figure 2 depicts our proposed
methodology.
1. Initial DW schema creation: involves the creation
of the initial DW schema following a classical
mixed approach;
2. NOSQL DW schema creation: it consists on the
generation of NOSQL data base for the initial data
warehouse schema based on a rules set. These
rules allow the transformation of a DW schema
concepts to specific concepts of NOSQL data
base;
3. Discovery of multidimensional concepts: consists
on defining the features set that meet the users
needs, generating clusters from social networks
based on the defined features and then determin-
ing the MC, their types as well as their locations;
4. NoSQL DW schema evolving: using the discov-
ered MC to enrich the NoSQL DW schema.
Figure 2: Overview of the proposed methodology.
In the following we expose the initial data ware-
house creation, then we detail the discovery of multi-
dimensional concepts using clustering technique.
4 INITIAL DW SCHEMA
CREATION: A DW FOR CRAFT
PRODUCTION ANALYSIS
In our research project, we need to create a DW to an-
alyze the impact of using IT solutions on the situation
of handicraft women. To accomplish this stage, we
have data sources (both internal and external) avail-
able in the project. They are the following:
S1: Data Base of the National Office for Tunisian
Handicrafts;
S2: Ontology defined from interviews;
S3: Tunisians and Algerian postcodes data bases.
In our case study, ontology (S2) represents an
internal source which is an essential component to
TowardsDataWarehouseSchemaDesignfromSocialNetworks-DynamicDiscoveryofMultidimensionalConcepts
341
evolve a DW schema. It contains both data and meta-
data. This source is a consistent support in the dis-
covery of dynamic multidimensional concepts. The
used ontology presents knowledge about the profile
of handicraft women. Similarly, the data Base of the
National Office for Tunisian Handicrafts (ONAT) is
an internal source. Otherwise, the Tunisian and Alge-
rian postcodes data bases s an external source which
is important to enrich the craft DW. The craft DW is
built from these three sources (S1, S2 and S3) using
the ETL (Extract Transform Load) process based on
data driven approach using Talend Open Source
5
. An
overall view of the initial craft DW schema is depicted
in Figure 3.
Figure 3: Data warehouse for craft production analysis.
The structure of the initial DW schema is a graph
centered at two facts. First, the Production fact: in-
cludes (Quantity, Pieces Number, Prod Duration)
as measures and is linked to (Artisan, Produc-
tion
Tool, Process, Raw Material, Product and
Date) as dimensions. Second, the Selling fact which
includes (Quantity, Amount) as measures and is
linked to (Artisan, Product and Date) as dimensions.
In this stage, we have followed a traditional approach
for DW schema design which is not suitable for the
latest data source (S4). Therefor, we propose to use
Clustering method.
5 DYNAMIC DISCOVERY OF
MULTIDIMENSIONAL
CONCEPTS
Classical decision support is used in analyzing sim-
5
http://www.talend.com/
ple data. However, these systems are not adapted
for SN analysis which highlights the need of creat-
ing new models. In fact, data are heterogeneous and
changeable over time. Thus, a comprehensive schema
for craft DW cannot be fixed at the time of design
and must be dynamically modified. In a nutshell,
a DW schema can be extended by adding new ele-
ments of type measure, dimension, or hierarchy level.
These extracted values are to be fed into the semiauto-
matic schema evolving to dynamically model the DW
schema. This stage is divided into three main steps:
features definition, clusters generation and multidi-
mensional concepts determination.
5.1 Features Definition
In DW lifecycle, user requirements definition is one
of the most important tasks which ensure a successful
DW project. The main objective is to identify analyst
goals in order to reduce the risk of failure. Since the
expert has prior knowledge of the analysis goal, our
methodology allows to evolve a DW schema under
the designers guidance. At first, the expert is required
to select the multidimensional concepts that should
be dynamically evolved according to his objectives.
He is then required to select one or a set of features
that meet his needs. These features are the basis for
grouping objects in the next step.
Example. To analyze the business of handicraft
women, we should cluster Products values based on
the Product Designation and then based on the used
Raw Material and the used Production Tool values.
In fact, if two products have the same designation,
they belong, therefore, to the same Group Activity.
Otherwise, if they use a set of Raw Material and Pro-
duction Tool in common, they probably belong to the
same Group Activity. Consequently, we derive three
features which are respectively Product Designation,
Raw Material and Production Tool.
5.2 Clusters Generation
The hierarchical clustering technique is applied to the
data set to generate clusters based on a similarity mea-
sure. As most of the clustering algorithms are un-
supervised, in this step, we target the semi super-
vised hierarchical clustering in order to get the op-
timal results that meet the analysts needs. To do
that, we have used the SHICARO (Semi-supervised
Hierarchical Clustering based on Ranking features
using Ontology) (R. Yangui and Gargouri, 2014a)
method with a profiling ontology. This method con-
sists of two important components. The first one
consists in defining effective ontology-based similar-
ICEIS2015-17thInternationalConferenceonEnterpriseInformationSystems
342
Figure 4: SHICARO method.
ity measures that combine both numerical and nom-
inal variables along different dimensions (instances,
attributes, and relation-ships) (R. Yangui and Gar-
gouri, 2014b), while the second consists in providing
a performable clustering algorithm based on ranking
features. SHICARO method aims to cluster objects
based on scheduled features.
Since the expert knows the goal behind which the
clustering is performed, SHICARO method performs
clustering under the users guidance. Thus, the user
is required to order the features from the highest to
the lowest ones. In each iteration, a set of features
F =
{
( f
1
, r
1
), . . . , ( f
n
, r
n
)
}
that have the same rank r
i
are applied to cluster objects. SHICARO steps are
depicted by Figure 4.
Example. By applying SHICARO method based
on the set of features F=(Product Designation,
1), (Raw Material, 2), (Production Tool, 2), we
obtained at the first iteration (based on Prod-
uct Designation feature) fives clusters and at the sec-
ond iteration (based on Raw Material and Produc-
tion Tool features) two clusters (Figure 5).
Figure 5: Clustering Product instances.
The extracted values, present in each of the gener-
ated clusters, become the input to the dynamic schema
evolving in the next step.
5.3 Multidimensional Concepts
Determination
This step is performed by the designer. It consists in
analyzing generated clusters, determining the type of
multidimensional concept, assigning names to clus-
ters and specifying the location of insertion in the
multidimensional schema.
Example1. Adding the Group Activity level to the
Product dimension hierarchy (Figure 6).
Figure 6: Adding the Group Activity level to the Product
dimension hierarchy.
In Figure 6.a, we have two clusters: the first one
describes the Traditional Weaving and the second
one describes the Traditional Clothing. So we can
define the concept Group Activity as MC. This MC is
considered as parameter of the Product dimension as
depicted in Figure 6.b.
Example2. Adding the Customer, Supplier and Fan
dimensions (Figure 7) Artisan friends in SN can be
divided into groups based on their job, their link and
the exchanged clips. Feature set used to cluster wom-
ens friends is F= (link, 1), (job,2), Clip,3). Clustering
algorithm based on F determine three groups named
Customer, Supplier and Fan.
TowardsDataWarehouseSchemaDesignfromSocialNetworks-DynamicDiscoveryofMultidimensionalConcepts
343
Figure 8: DW schema enriched by discovered MC.
Figure 7: Adding the Customer, Supplier and Fan dimen-
sions.
Each group represents a CM, so we have three
new MC: Customer, Supplier and FAN. These MC
are considered as three new dimensions. Customer is
added as dimensions to the Selling fact. Supplier and
FAN are added as dimensions to the Production fact
as depicted in figure 7.
Figure 8 shows the DW schema enrichment for
storing various cumulative data about handicraft
women derived from SN in the the extended Dimen-
sional Fact Model (xDFM) (Mansmann, 2008). Di-
mensions are modeled as aggregation paths. All paths
of a dimension converge in an abstract Tnode, which
corresponds to the aggregated value all. A level node
in a dimension consists of at least one key attribute,
but may include further attributes represented as un-
derlined terminal nodes. As shown in Figure 8, three
new dimensions are added to the initial DW schemas
which are Customer, Supplier and Fan. These di-
mensions are discovered by performing the hierar-
chical clustering based on a set of features. Cus-
tomer, Supplier and Fan reflect the communities to
which handicraft women are connected at SN. How-
ever, an Artisan can be also a Supplier, a Customer
or a Fan. That is why a generalization link is defined
Tperson. Similarly, Raw Mat Type, Activity Group
and Tool Type proprieties are added as levels succes-
sively in the Raw Material, Product and Fan dimen-
sion hierarchies. These properties are deducted us-
ing the hierarchical clustering based on the appro-
priate features. However, a Product can also be a
Raw Material. Hence, a generalization link is added
Tmaterial.
6 CONCLUSION
In this paper, we proposed four stage methodology
to define a data warehouse schema from social net-
works. Starting in a first stage, by the design of the
initial data warehouse schema based on the existing
approaches. In a second stage, we have generated a
NOSQL Data warehouse schema by applying a set of
transformation rules. Then, based on users require-
ICEIS2015-17thInternationalConferenceonEnterpriseInformationSystems
344
ment, clustering of social networks profiling data is
performed which allows the dynamic discovery of
multidimensional concepts. Finally, the enrichment
of the NoSQL data warehouse schema by the discov-
ered MC is realized to ensure the DW schema evolu-
tion.
We have especially detailed the dynamic content-
based discovery of dimensions, hierarchies and mea-
sures using hierarchical clustering. This latter, is per-
formed using profiling ontology with adequate simi-
larity measures. The detailed stages are experimented
on the real case study of the BWEC project.
We are currently studying the NOSQL data bases,
we intend to define transformation rules from a con-
ceptual data warehouse schema to NOSQL database
and the evolution rules. Moreover, we think it would
be interesting to formally specifying transformation
rules to allow the automatic schema generation.
ACKNOWLEDGEMENT
We are very thankful to the Algerian Tunisian Project
dealing with the improvement of handicraft women
business in emerging countries through affordable
technologies and social networks. This project is fi-
nanced by the Tunisian Ministry of Higher Education,
Scientific Research and Information and Communi-
cation Technologies Higher Education and Scientific
Research sector.
REFERENCES
A. Thusoo, Z. Shao, S. A. D. B. N. J. J. S. S. R. M. and Liu,
H. (2010). Data warehousing and analytics infrastruc-
ture at facebook. In International conference on Man-
agement of data SIGMOD’10, pages 1013–1020.
E. Gallinucci, M. Golfarelli, A. W. and Rizzi, S. (2013).
Meta-stars: Multidimensional modeling for social
business intelligence. In International Workshop On
Data Warehousing and OLAP DOLAP13, pages 11–
18.
L. Sautot, B. Faivre, L. J. and Molin, P. (2014). The hi-
erarchical agglomerative clustering with gower index:
A methodology for automatic design of olap cube in
ecological data processing context. In Ecological In-
formatics, pages 1–14.
M. Ceci, A. and Malerba, D. (2011). Olap over contin-
uous domains via densitybased hierarchical cluster-
ing. In International Conference on Knowledge-Based
and Intelligent Information and Engineering Systems
KES’11, pages 559–570.
Mansmann, S. (2008). Extending the olap technology
to handle nonconventional and complex data. In
Ph.D. dissertation, University Konstanz, Department
of Computer and Information Science Germany.
Moalla, I. and Nabli, A. (2014). Towards data mart build-
ing from social network for opinion analysis. In Inter-
national Conference on Intelligent Data Engineering
and Automated Learning IDEAL14, pages 295–302.
N. U. Rehman, S. Mansmann, A. W. and Scholl, M. H.
(2012). Building a data warehouse for twitter stream
exploration. In International Conference on Advances
in Social Networks Analysis and Mining ASONAM’12,
pages 1341–1348.
Nabli, A. (2013). Approche d’aide la conception au-
tomatise d’entrept de donnes: Guide de modlisation.
Presses Acadmiques Francophones.
P. Kazienko, K. Musial, E. K. T. K. and Brdka, P. (2011).
Multidimensional social network: Model and analy-
sis. In International Conference on Computational
Collective Intelligence Technologies and Applications
ICCCI’11, pages 378–387.
R. Yangui, A. N. and Gargouri, F. (2014a). Shicaro: Semi-
supervised hierarchical clustering based on ranking
features using ontology. In International Conference
on Management and Technology in Knowledge, Ser-
vice, Tourism & Hospitality SERVE’14, pages 233–
238.
R. Yangui, A. N. and Gargouri, F. (2014b). Soim: Simi-
larity measures on ontology instances based on mixed
features. In 4th International Conference on Model &
Data Engineering MEDI2014, pages 169–176.
Usman, M. and Pears, R. (2011). Multi level mining of
warehouse schema. In NDT, volume 136 of Communi-
cations in Computer and Information Science, pages
395–408.
TowardsDataWarehouseSchemaDesignfromSocialNetworks-DynamicDiscoveryofMultidimensionalConcepts
345