A UML-EXTENDED APPROACH FOR MINING OLAP DATA
CUBES IN COMPLEX KNOWLEDGE DISCOVERY
ENVIRONMENTS
Alfredo Cuzzocrea
ICAR-CNR and University of Calabria, Cosenza, Italy
Keywords: Mining OLAP data cubes, UML, Complex knowledge discovery environments.
Abstract: In this paper, we propose theoretical assertions and practical instances of an innovative UML-extended
approach for mining OLAP data cubes in complex knowledge discovery environments. This analytical
contribution is further extended by means of a comprehensive set of case studies that clearly demonstrate
the feasibility and the benefits of the proposed approach in the context of next generation Data-
Warehousing/Data-Mining platforms.
1 INTRODUCTION
The problem of mining OLAP data cubes (e.g.,
Inmon, 1996; Han, 1997; Cuzzocrea, 2009) plays a
leading role in next-generation complex Data
Mining systems and applications such as social
networks, analytics over very large distributed
repositories, stream and sensor data analysis, and so
forth. This maily because of data repositories stored,
processed and mined in real-life systems and
applications are ineherently multidimensional, multi-
level and multi-resolution in nature (Cai et al., 2004;
Han et al., 2005; Cuzzocrea et al., 2009) hence
executing Data Mining algorithms over
multidimensional views computed on top of original
data sources leads to more expressive and powerful
mining results (Inmon, 1996; Han, 1997; Rizzi et al,
2006; Cuzzocrea, 2007; Cuzzocrea, 2009).
The main idea of mining-OLAP-data-cube
models and algorithms consists in formulating novel
versions of traditional Data Mining approachs over
flat data sources (e.g., relational databases) (Frawley
et al., 1992; Fayyad et al., 1996) hence stirring-up
towards innovative achievements that fully consider
the multidimensionality and the multi-resolution of
data and schema cubes (Gray et al., 1997).
This conceptual setting becomes more and more
difficult in complex knowledge discovery
environments, where knowledge discovery processes
and routines are subjected to complex constraints
and external conditions, and algorithms naturally
become harder and harder as well. Relevant
instances of such environments are stream and
sensor data analysys tools, as mining stream and
sensor data (Gaber at al., 2005) is a very challenging
task due to a plethora of aspects ranging from
unbounded length of streams to the need for single-
pass (mining) algorithms, from multi-rate arrivals to
uncertainity and imprecision, and so forth.
Another important aspect that adds complexity in
mining OLAP data cubes is represented by the fact
that, as recognized in (Zubcoff & Trujillo, 2006;
Zubcoff & Trujillo, 2007; Zubcoff et al., 2007a;
Zubcoff et al., 2007b), Data Mining is still an
artifact-like and solution-driven task, which heavily
depends on the particular application context. In
practice, given a certain Data Mining problem, each
data miner is likely to develop a proper Data Mining
solution (with proper Data Mining models and
algorithms), by referring-to and extending state-of-
the-art results. Also, the solution usually turns to be
very different from solutions proposed by other data
miners. While this is reasonable, it has been
recognized that solution-driven approaches
seriously limit the integration of software
engineering paradigms (Heineman & Councill,
2001), like conceptual modeling, project-sharing, re-
engineering strategies, and so forth, within Data
Mining (Zubcoff & Trujillo, 2006; Zubcoff &
Trujillo, 2007; Zubcoff et al., 2007a; Zubcoff et al.,
281
Cuzzocrea A..
A UML-EXTENDED APPROACH FOR MINING OLAP DATA CUBES IN COMPLEX KNOWLEDGE DISCOVERY ENVIRONMENTS.
DOI: 10.5220/0003512302810289
In Proceedings of the 13th International Conference on Enterprise Information Systems (ICEIS-2011), pages 281-289
ISBN: 978-989-8425-53-9
Copyright
c
 2011 SCITEPRESS (Science and Technology Publications, Lda.)

2007b). Furthermore, while this is critical in
conventional Data Mining settings, it plays an even
more challenging role in mining multidimensional
structures like OLAP data cubes stored in Data
Warehosue architectures (Zubcoff & Trujillo, 2006;
Zubcoff & Trujillo, 2007; Zubcoff et al., 2007a;
Zubcoff et al., 2007b), due to the complexities
introduced by multidimensional data and schemas.
Starting from these considerations, in (Cuzzocrea
et al., 2011) an innovative model-driven Data
Mining engineering framework for moving Data
Mining models/algorithms from solution-oriented
artifacts to “composable” Data Mining conceptual
models has been proposed. This framework allows
us to enable more sophisticated theoretical settings
leading to software-engineering-inspired Data
Mining approaches, hence fully incorporating well-
understood software-component design paradigms
(Heineman & Councill, 2001).
A particularity of this framework is represented
by the fact that it completely focuses on
multidimensional data structures, due the above-
mentioned benefits deriving from mining
multidimensional cubes and views computed over
the original data sets with respect to the more
conventional, baseline case of mining flat data
sources (Inmon, 1996; Han, 1997; Rizzi et al, 2006;
Cuzzocrea, 2007; Cuzzocrea, 2009).
Basically, the main idea underlying the
framework (Cuzzocrea et al., 2011) consists in
modeling both the multidimensional data
reportiories and the mining algorithms by means of
innovative UML profiles that allow us to build
conceptual models rather than solution-driven
impelmtations. This allow analysts to concentrate on
the abstract and conceptual level of the target Data
Mining problem, rather than wasting time on low-
level aspects. Notice that, in complex knowledge
discovery environments, this feature plays a leading
role, as scalability (even at a conceptual/modeling)
becomes a challenge in such environments. Another
important characteristic of the framework proposed
in (Cuzzocrea et al., 2011) consists in the fact that a
set of suitable model transformations is able of
generating both the data under analysis (in an
apposite Data Warehouse platform) and the analysis
model (in a proper Data Mining platform). Indeed, it
should be clear enough that this approach keeps
several points of research innovation in the context
of Data-Warehousing/Data-Mining research.
This paper significantly extends the research
results provided in (Cuzzocrea et al., 2011), and
provides a a comprehensive set of case studies
focused on the application of well-understood and
popular Data Mining techniques to a large corporate
database of a hypothetical electric-supply company,
where we apply the model-driven Data Mining
engineering approach (Cuzzocrea et al., 2011). The
final goal is that of demonstrating and validating the
feasibility of this approach throughout significant
examples, while also clarifying theoretical and
methodological aspects correlated to the research
results provided in (Cuzzocrea et al., 2011). The
Data-Mining/Analysis platform taken as reference
for platform-dependent implementations generated
by our model-driven Data Mining engineering
approach (see Section 4) is MS Analysis Services
(Microsoft Research, 2009b), which makes use of
Data Mining eXtensions (DMX) (Microsoft
Research, 2009a) as mining/modeling
language/formalism (see Section 3).
With these goals in mind, the rest of the paper is
structured as follows. In Section 2, we describe the
(large) corporate database of the hypothetical
electric-supply company, by focusing the attention
on most-distinctive characteristics.
We then move the attention on the application of
two well-understood and popular Data Mining
techniques over the electric-supply corporate
database via following the methodological
guidelines drawn by the model-driven Data Mining
engineering framework (Cuzzocrea et al., 2011).
This also implies, as a secondary task, the definition
of a collection of suitable OLAP data cube models
that aggregate multidimensional data within apposite
multidimensional data cubes able of providing
support for integrated, cleaned and aggregated
multidimensional views over which appropriate Data
Mining techniques can be applied with success
(more than the case for traditional database-oriented
conceptual KDD reference architecture – (Inmon,
1996; Han, 1997; Rizzi et al, 2006; Cuzzocrea,
2007; Cuzzocrea, 2009)). For each of these Data
Mining techniques, we provide the platform-
dependent implementation over MS Analysis
Services as generated by the approach (Cuzzocrea et
al., 2011), plus analysis and discussion of (mining)
results still in the same platform. The Data Mining
techniques we consider in this paper are the
following: clustering (Tan et al., 2005a; Tan et al.,
2005b), which is presented in Section 3, and
decision trees (Quinlan, 1986), which is presented in
Section 4.
ICEIS 2011 - 13th International Conference on Enterprise Information Systems
282

Figure 1: Schema of the electric-supply corporate database used as reference case study.
After that, in Section 5 we prove the component-
oriented capabilities of the model-driven Data
Mining engineering approach (Cuzzocrea et al.,
2011) via combining the so-obtained clustering
conceptual mining model with the so-obtained
decision-trees conceptual mining model into a novel
clustering /decision-trees mining model, with the
goal of enhancing the expressive power and the
effectiveness of the overall Data Mining modeling
process over the electric-supply corporate database.
Finally, in Section 6 we provide conclusions of
the research presented in this paper and future
research directions to be further investigated.
2 THE ELECTRIC-SUPPLY
CORPORATE DATABASE
Figure 1 shows the schema of the electric-supply
corporate database used as reference data source in
the in-laboratory validation of the model-driven
Data Mining engineering approach (Cuzzocrea et al.,
2011). The electric-supply corporate database stores
all the information needed to support the activities of
the electric-supply company, such as: (i) information
needed to handle the main electric-supply network
managed by the company; (ii) information needed to
monitor and operate on the national electric-supply
market; (iii) information needed to manage users of
the company; (iv) information needed to manage the
electric-supply distribution network across the
country.
In the following, we provide an overview of the
main tables populating the electric-supply corporate
database:
1. GRTN: it stores data related to the handling of
the main electric-supply network managed by
the company. Most significant attributes are the
following: (i) GRTN_NodeID, which models the
absolute identifier of the actual electric-supply
network node; (ii) Sector, which models the
network sector to which the actual node
belongs-to; (iii) SharedCapital, which models
the amount of electric-supply power the actual
node manages; (iv) Address, which models the
address where the actual node is located.
2. GME: it stores data related to the monitoring
and the operating on the national electric-supply
market. Most significant attributes are the
following: (i) GME_ NodeID, which models the
absolute identifier of that node whit respect to
which supply and demand of the electric-supply
market are managed; (ii) Sector, which models
the network sector to which the actual node
belongs-to; (iii) SharedCapital, which models
the amount of electric-supply power the actual
node manages; (iv) Address, which models the
address where the actual node is located.
3. User: it stores data related to the management of
users of the company. Most significant
attributes are the following: (i) UserID, which
models the absolute identifier of the actual user
of the electric-supply company; (ii) Type, which
models the class of the actual user; (iii)
ElectricityMeter, which models the absolute
identifier of the electricity meter in usage by the
actual user; (iv
) FirstName, which models the
first name of the actual user; (v) LastName,
which models the first name of the actual user;
(vi) Address, which models the address of the
A UML-EXTENDED APPROACH FOR MINING OLAP DATA CUBES IN COMPLEX KNOWLEDGE DISCOVERY
ENVIRONMENTS
283

actual user; (vii) Phone, which models the
phone number of the actual user.
4. DistributionDepartment: it stores data related to
the management of the electric-supply
distribution network across the country. Most
significant attributes are the following: (i)
DistDeptID, which models the absolute
identifier of the actual distribution department;
(ii) ProductionCapacity, which models an
indicator representing the production capacity of
the actual department; (iii) SharedCapital,
which models the amount of electric-supply
power the actual department produces; (iv)
CallCenter, which models the absolute
identifier of the call center associated with the
actual department.
3 THE CLUSTERING
CONCEPTUAL DATA MINING
MODEL USERCLUSTERING
The first case study of our analysis focuses on the
application of a clustering technique (Tan et al.,
2005a; Tan et al., 2005b) over user data of the
hypothetical electric-supply company, which we
name as UserClustering, whose final goal is that of
clustering users based on their electric-supply
purchases (represented in terms of billing
documents). Figure 2 shows the conceptual Data
Mining model of UserClustering shaped according
to the model-driven Data Mining engineering
approach (Cuzzocrea et al., 2011). As shown in
Figure 2, the data layer of UserClustering is
captured by the three-dimensional OLAP data cube
Sales, whereas Expectation Maximization (EM)
(Dempster et al., 1977) has been adopted as specific
clustering algorithm. Looking into detail, Sales is
characterized by the fact Document, which models
information about the billing documents, and by the
following dimensions: (i) User, which models users
of the electric-supply company; (ii) CallCenter,
which models call centers of the electric-supply
company; (iii) Date, which is a standard (OLAP)
temporal dimension. For what instead regards the
algorithmic parameters of the target EM clustering
algorithm, the minimum support has been set as
equal to 1 and the maximum number of generated
clusters has been set as equal to 10, respectively.
Attributes Address and Type of the dimension User,
and attributes Tax and Taxable of the fact Document,
respectively, have been set as input model
parameters of the EM clustering algorithm, whereas
User has been set as case dimension (i.e., a
dimension containing case parameters – see Section
4.2) for UserClustering.
Figure 2: Conceptual Data Mining model of
UserClustering.
We now move the attention on the platform-
dependent implementations of UserClustering
generated by the model-driven Data Mining
engineering approach (Cuzzocrea et al., 2011) on
MS Analysis Services (Microsoft Research, 2009b).
To this end, Figure 3 shows the implementation of
the data cube Sales. Figure 4 shows the mining
structure of UserClustering (i.e., the set of data and
meta-data supporting the execution of
UserClustering within the core layer of MS Analysis
Services). Finally, Figure 5 shows the final platform-
specific realization of UserClustering and Figure 6
shows the corresponding DMX-based
implementation, respectively.
Figure 3: The data cube Sales of UserClustering.
ICEIS 2011 - 13th International Conference on Enterprise Information Systems
284

Figure 4: The mining structure of UserClustering.
Figure 5: The final realization of UserClustering.
CREATE MINING MODEL
UserClustering{
UserID text key,
Type text discrete,
Taxable long continuous,
Tax long continuous,
Address text discrete
} USING Microsoft_Clustering(
MINIMUM_SUPPORT = 1, CLUSTER_COUNT = 10);
Figure 6: The DMX-based implementation of
UserClustering.
As highlighted above, in our analysis we again
make use of the underlying Data-Mining/Analysis
platform
(MS Analysis Services, in this case) to
provide discussion of retrieved (mining) results
(clusters, in this case). To this end, Figure 7 shows
the results obtained from executing UserClustering
over the target multidimensional data repository.
Here, nodes represent clusters, arcs represent
relations
between clusters (e.g., hierarchical
Relations), and darker clusters represent clusters
grouping higher-in-cardinality user populations.
Figure 7: Mining results of UserClustering.
4 THE DECISION-TREES
CONCEPTUAL DATA MINING
MODEL
MAINTENANCECAUSAL
The second case study of our analysis deals with the
application of a decision-tree technique (Quinlan,
1986) over service data of the electric-supply
company, named as MaintenanceCausal.
MaintenanceCausal aims at discovering major
causal of maintenance services performed in the
main electric-supply network managed by the
company. Figure 8 shows the conceptual Data
Mining model of this second case study. As shown
in Figure 8, the three-dimensional OLAP data cube
Planning
characterizes the data layer of
MaintenanceCausal. In more detail, Planning
introduces the fact Service, which models
information about the services performed within the
Figure 8: Conceptual Data Mining model of
MaintenanceCausal.
A UML-EXTENDED APPROACH FOR MINING OLAP DATA CUBES IN COMPLEX KNOWLEDGE DISCOVERY
ENVIRONMENTS
285

electric-supply company, and the following
dimensions: (i) Maintenance, which models
information about specific maintenance operations
of (maintenance) services; (ii) Platform, which
models information about the platform where
services are performed – Platform also comprises a
normalized dimensional table,
ProductionDepartment; (iii) ProductionData, which
is a standard (OLAP) temporal dimension. For what
instead regards the specific decision-trees algorithm,
MaintenanceCausal makes use of the ad-hoc
implementation provided by MS Analysis Services
that consists of a hybrid decision-tree algorithm
(Microsoft Research, 2009b) able of supporting both
classification and regression functionalities during
the mining phase. For this algorithm, the minimum
support has been set as equal to 10 and the
partitioning method has been chosen as the binary
one, respectively. Furthermore, attributes Quantity
and Price of the fact Service, and Hour, Description,
Type and Material of the dimension Maintenance,
respectively, have been set as input model
parameters of the decision-trees algorithm, whereas
attribute Causal of the dimension Maintenance has
been set as output model parameter.
Figure 9 shows the platform-dependent
implementation of the data cube Planning. Figure 10
shows instead the mining structure of
MaintenanceCausal, whereas, to conclude, Figure
11 shows the final realization of MaintenanceCausal
and Figure 12 shows the corresponding DMX-based
implementation, respectively.
Figure 9: The data cube Planning of MaintenanceCausal.
Figure 10: The mining structure of MaintenanceCausal.
Figure 11: The final realization of MaintenanceCausal.
CREATE MINING MODEL
MaintenanceCausal{
MaintenanceID long key,
Causal text discrete predict_only,
Description text discrete,
Hours date continuous,
Material text discrete,
Price long continuous,
Quantity long continuous,
Type text discrete
} USING Microsoft_Decision_Trees(
MINIMUM_SUPPORT = 10, SPLIT_METHOD = 2);
Figure 12: The DMX-based implementation of
MaintenanceCausal.
Results obtained from executing
MaintenanceCausal over the target
multidimensional data repository are depicted in
Figure 13. Here, darker nodes represent decision-
tree nodes classifying higher counts of maintenance
services,
and the histogram stored by each decision-
tree node captures the distribution of maintenance
ICEIS 2011 - 13th International Conference on Enterprise Information Systems
286

Figure 13: Mining results of MaintenanceCausal.
services with respect to the following causal: (i)
“material breaches”, represented by a blue bar; (ii)
“technical assistance”, represented by a red bar;, (iii)
“ordinary assistance”, represented by a green bar.
5 COMBINING CONCEPTUAL
DATA MINING MODELS
A nice amenity supported by the model-driven Data
Mining engineering approach (Cuzzocrea et al.,
2011) consists in the fact that obtained conceptual
Data Mining models can be further combined into a
novel model, according to widely-understood
software-component design paradigms (Heineman &
Councill, 2001). This with the goal of enhancing the
expressive power of the overall Data Mining
modeling process. In line with this leading aspect, in
the third (and last) case study we combine the
clustering conceptual Data Mining model
UserClustering and the decision-trees conceptual
Data Mining model MaintenanceCausal, which have
been presented and discussed in Section 3 and
Section 4, respectively. This allows us to obtain a
novel clustering/decision-trees mining model, named
as UserClusteredMaintenanceCausal, which is
depicted in Figure 14.
UserClusteredMaintenanceCausal is a complex
conceptual Data Mining model whose main goal
consists in discovering major causal of maintenance
services performed in the main electric-supply
network managed by the company (which is the
specific goal of MaintenanceCausal – see Section 4)
by grouping per clusters of users generated on the
basis of users’ electric-supply purchases (which is
the specific goal of UserClustering – see Section 3).
It should be noted that the Data Mining analysis
supported by UserClusteredMaintenanceCausal
turns to be extremely useful for decision makers of
the hypothetical electric-supply company, as, based
on results derived from executing
Figure 14: Conceptual Data Mining model of
UserClusteredMaintenanceCausal.
UserClusteredMaintenanceCausal over suitable
repositories of multidimensional data, decision
makers can determine electric-supply distribution
policies different from the actual ones if clear
relations among users’ electric-supply purchases and
maintenance service causal are discovered (e.g.,
peaks of users’ electric-supply purchases may
frequently involve in high numbers of request for
technical assistance).
As shown in Figure 14, in
UserClusteredMaintenanceCausal the output
provided by UserClustering (i.e., user clusters) is
given as input to MaintenanceCausal, and the final
result is still modeled in terms of a conventional
decision tree, like in MaintenanceCausal. In
particular, UserClustering is used like a sort of novel
“dimension” of the three-dimensional OLAP data
cube Sales, which originally was part of
UserClustering (see Section 3), and the artificial
attribute NODE_UNIQUE_NAME, which models a
node-based representation of user clusters obtained
from UserClustering (like the one shown in Figure
7), has been set as input model parameter for
MaintenanceCausal.
Figure 15 shows the mining structure of
UserClusteredMaintenanceCausal
(recall that the
A UML-EXTENDED APPROACH FOR MINING OLAP DATA CUBES IN COMPLEX KNOWLEDGE DISCOVERY
ENVIRONMENTS
287

Figure 15: The mining structure of
UserClusteredMaintenanceCausal.
implementation of Sales if the same of the one
provided for UserClustering, which is depicted in
Figure 3). Furthermore, Figure 16 shows the final
realization of UserClusteredMaintenanceCausal and
Figure 17 shows the corresponding DMX-based
implementation, respectively.
Results obtained from executing
UserClusteredMaintenanceCausal over the target
multidimensional data repository are shown in
Figure 18 (due to space reasons, only the first four
UserClustering (see Figure 13), darker nodes
represent decision-tree nodes classifying higher
levels
of the output decision tree are shown). Here,
like for the case of the mining results of counts of
maintenance services grouped by user clusters, and
the histogram stored by each decision- tree node
captures the distribution of maintenance services
with respect to the following user clusters: (i) “home
users”, represented by a blue bar; (ii) “partner
users”, represented by a red bar; (iii) “Business
users”, represented by a green bar.
Figure 16: The final realization of
UserClusteredMaintenanceCausal.
CREATE MINING MODEL
UserType{
UserID long key,
Type text discrete predict_only,
UserClustering_DMDim table (
NODE_UNIQUE_NAME text key)
} USING Microsoft_Decision_Trees(
MINIMUM_SUPPORT = 10, SPLIT_METHOD = 2);
Figure 17: The DMX-based implementation of
UserClusteredMaintenanceCausal.
6 CONCLUSIONS AND FUTURE
WORK
Starting from the research results provided in
(Cuzzocrea et al., 2011), where an innovative
model-driven Data Mining engineering framework
that focuses on multidimensional data structures has
been proposed, in this paper we have further and
significantly extended these results by proposing a
comprehensive set of case studies focused on the
application of well-understood and popular Data
Mining techniques to a large corporate database of a
hypothetical electric-supply company. These case
studies are meant to prove some relevant
characteristics of the framework (Cuzzocrea et al.,
2011), among which the suitability of this
framework in providing conceptual Data Mining the
Figure 18: Mining results of UserClusteredMaintenanceCausal.
ICEIS 2011 - 13th International Conference on Enterprise Information Systems
288

nice amenity of making these model “composable”
in order to achieve the definition of models for Data
Mining problems arising in complex knowledge
discovery environments, and complex Data Mining
models starting from simpler ones play the major
roles.
Future work of this research is oriented towards
integrating within the framework (Cuzzocrea et al.,
2011) innovative aspects as to capture advanced
features of Data-Warehouse/Data-Mining platforms,
such as security and privacy, and uncertainty and
imprecision.
REFERENCES
Cai, Y. D., Clutterx, D., Papex, G., Han, J., Welgex, M.,
and Auvilx, L. (2004) ‘MAIDS: mining alarming
incidents from data streams’, in Proceedings of the
2004 ACM International Conference on Management
of Data, pages 919–920.
Cuzzocrea, A. (2007) ‘An OLAM-based framework for
complex knowledge pattern discovery in distributed-
and-heterogeneous-data-sources and cooperative
information systems’, in Proceedings of the 9th
International Conference on Data Warehousing and
Knowledge Discovery, LNCS Vol. 4654, pages 181–
198.
Cuzzocrea, A. (2009) ‘OLAP intelligence: meaningfully
coupling OLAP and data mining tools and
algorithms’, International Journal of Business
Intelligence and Data Mining, 4(3-4): 213–218.
Cuzzocrea, A. Furfaro, F., Masciari, E., and Saccà D.
(2009) ‘Improving OLAP analysis of
multidimensional data streams via efficient
compression techniques’, in A. Cuzzocrea (ed.),
“Intelligent Techniques for Warehousing and Mining
Sensor Network Data”, IGI Global, 17–49.
Cuzzocrea, A., Mazon, J.-N., Trujillo J., and Zubcoff, J.
(2011) ‘Model-driven data mining engineering: from
solution-driven implementations to “composable”
conceptual data mining models”, International
Journal of Data Mining, Modelling and Management,
to appear.
Dempster, A. P., Laird, N. M., and Rubin, D. B. (1977)
‘Maximum likelihood from incomplete data via the
EM algorithm’, Journal of the Royal Statistical
Society, Series B, 39(1):1–38.
Frawley, W., Piatetsky-Shapiro, G., and Matheus, C.
(1992) ‘Knowledge discovery in databases: an
overview’, AI Magazine, 13(3): 213–228.
Fayyad, U., Piatetsky-Shapiro, G., Smyth, P., and
Widener, T. (1996) ‘The KDD process for extracting
useful knowledge from volumes of data’,
Communications of the ACM, 39(11):27–34.
Gaber, M. M., Zaslavsky, A. B., and Krishnaswamy, S.
(2005) ‘Mining data streams: a review’, SIGMOD
Record, 34(2): 18–26.
Gray, J., Chaudhuri, S., Bosworth, A., Layman, A.,
Reichart, D., Venkatrao, M., Pellow, F., and Pirahesh,
H. (1997) ‘Data cube: a relational aggregation
operator generalizing group-by, cross-tab, and sub-
totals’, Data Mining and Knowledge Discovery,
1(1):29–53.
Han, J. (1997) ‘OLAP mining: an integration of OLAP
with data mining’, in Proceedings of the 7th IFIP 2.6
Working Conference on Database Semantics, pages 1–
9.
Han, J., Chen, Y., Dong, G., Pei, J., Wah, B. W., Wang, J.,
and Cai, Y. D. (2005) ‘Stream cube: an architecture
for multi-dimensional analysis of data streams’,
Distributed and Parallel Databases, 18(2): 173–197.
Heineman, G. T., and Councill, W. T. (2001) Component-
Based Software Engineering: Putting the Pieces
Together, Addison-Wesley Professional, Reading,
MA, USA.
Inmon, W. H. (1996) ‘The data warehouse and data
mining’, Communications of the ACM, 49(4)83–88.
Microsoft Research (2009a) Data Mining eXtensions
(DMX) Reference, http://msdn.microsoft.com/en-us/
library/ ms132058.aspx
Microsoft Research (2009b) SQL Server Analysis Services
– Data Mining, http://msdn.microsoft.com/en-us/
library/ bb510517.aspxQuinlan, J.R. (1986) ‘Induction
of decision trees’, Machine Learning, 1(1):81–106.
Tan, P.-N., Steinbach, M., and Kumar, V. (2005a)
Introduction to data mining – Chapter 8: Cluster
analysis: basic concepts and algorithms, Addison-
Wesley, Reading, MA, USA.
Tan, P.-N., Steinbach, M., and Kumar, V. (2005b)
Introduction to data mining – Chapter 9: Cluster
analysis: additional issues and algorithms, Addison-
Wesley, Reading, MA, USA.
Zubcoff, J., and Trujillo, J. (2006) ‘Conceptual modeling
for classification mining in data warehouses’, in
Proceedings of the 8th International Conference on
Data Warehousing and Knowledge Discovery, LNCS
Vol. 4081, pages 566–575.
Zubcoff, J., and Trujillo, J. (2007) ‘A UML 2.0 profile to
design association rule mining models in the
multidimensional conceptual modeling of data
warehouses’, Data & Knowledge Engineering,
63(1):44–62.
Zubcoff, J., Pardillo, J., and Trujillo, J. (2007a)
‘Integrating clustering data mining into the
multidimensional modeling of data warehouses with
UML profiles’, in Proceedings of the 9th International
Conference on Data Warehousing and Knowledge
Discovery, LNCS Vol. 4654, pages 199–208.
Zubcoff, J., Trujillo, J., and Cuzzocrea, A. (2007b) ‘On
the suitability of time series analysis on data
warehouses’, in Proceedings of the 1st IADIS
European Conference on Data Mining, pages 17–24.
A UML-EXTENDED APPROACH FOR MINING OLAP DATA CUBES IN COMPLEX KNOWLEDGE DISCOVERY
ENVIRONMENTS
289