=Paper=
{{Paper
|id=Vol-2236/paper-08-008
|storemode=property
|title=
Objects of Interest Detection by Earth Remote Sensing Data Analysis
|pdfUrl=https://ceur-ws.org/Vol-2236/paper-08-008.pdf
|volume=Vol-2236
|authors=Anastasia V. Demidova,Maxim B. Fomin,Sergey G. Shorokhov
}}
==
Objects of Interest Detection by Earth Remote Sensing Data Analysis
==
https://ceur-ws.org/Vol-2236/paper-08-008.pdf
65
UDC 681.3.016
Objects of Interest Detection by Earth Remote Sensing Data
Analysis
Anastasia V. Demidova* , Maxim B. Fomin† , Sergey G. Shorokhov †
*
Department of Applied Probability and Informatics,
Peoples’ Friendship University of Russia (RUDN University),
6 Miklukho-Maklaya str., Moscow, 117198, Russian Federation
†
Department of Information Technologies
Peoples’ Friendship University of Russia (RUDN University)
6 Miklukho-Maklaya str., Moscow, 117198, Russian Federation
Email: demidova-av@rudn.ru, fomin_mb@rudn.university, shorokhov_sg@rudn.university
In information systems based on a multidimensional approach, a data model is a multidi-
mensional data cube. If one uses a large set of aspects for the analysis of data domain the
data cubes are characterized by substantial sparseness. This makes it difficult to describe the
metadata of the information system and complicates the organization of data storage. To
describe the structure of a sparse data cube, a cluster method can be used. This method is
based on the construction of groups of members which are semantically connected with other
groups of members. Connected groups related to different dimensions describe the cluster of
cells. Classification schemes that correspond to the structural components of the observed
phenomenon can be used to describe it’s semantics. Every classification scheme is a graph de-
scribing the hierarchy of members that are associated with a separate structural component
of the observed phenomenon. The coupling between several classification schemes related to
different structural components helps to describe the metadata of the multidimensional infor-
mation system. Classification schemes are a source of classification of information objects of
a multidimensional cube related to the structural components of the observed phenomenon.
Key words and phrases: OLAP, data warehouse, multidimensional data model, sparse
data cube, set of possible member combinations, cluster of member combinations.
Copyright © 2018 for the individual papers by the papers’ authors. Copying permitted for private
and academic purposes. This volume is published and copyrighted by its editors.
In: K. E. Samouylov, L. A. Sevastianov, D. S. Kulyabov (eds.): Selected Papers of the 1st Workshop
(Summer Session) in the framework of the Conference “Information and Telecommunication
Technologies and Mathematical Modeling of High-Tech Systems”, Tampere, Finland, 20–23 August,
2018, published at http://ceur-ws.org
66 ITTMM-WSS—2018
1. Introduction
Multidimensional information systems based on the principles of OLAP are used
for the operational analysis of large datasets. Analytical space in a system of this
type is a multidimensional data cube. The role of the cube dimensionalities is played
by the dimensions corresponding to various aspects of the observed phenomenon for
which description the system is developed. If we use a large amount of semantically
heterogeneous data for the description of the observed phenomenon the multidimensional
cube is characterized by high sparseness and irregular filling [1–8]. As a result, there is a
problem of developing an adequate way to describe the structure of an analytical space
which use would make it possible to effectively organize the data analysis process [9–
17]. Such a correct way should provide the accounting of semantics of the observed
phenomenon.
The cluster method can be used for the effective description of the multidimensional
cube structure. This method is based on the semantic analysis of different dimensions’
members’ compatibility in possible cube cells. It allows describing the metadata of the
information system as a set of possible member combinations. Possible combinations
comply with possible cells of the multidimensional cube. Every possible cube cell
complies with some fact.
Difficulties in describing the structure of the analytical space may arise in case if,
in the process of forming the metadata of the information system, the analysis of the
semantic aspects of the observed phenomenon is subject to technological aspects. The
observed phenomenon is a set of interrelated processes related to the subject domain.
Data describing the observed phenomenon can form one or more multidimensional data
cubes. In describing the structure of an analytic space, the following problems must be
solved:
– the problem of classification of data describing the observed phenomenon;
– the problem of accounting for the semantics of the observed phenomenon in these
data.
2. Multidimensional data model
The structure of multidimensional data model should reflect the aspects of subject
domain which are used in the data analysis process. Each aspect corresponds to
one dimension
{︀ of a multidimensional
}︀ cube 𝐻. A full set of dimensions forms a set
𝐷(𝐻) = 𝐷1 , 𝐷2 , .., 𝐷𝑛 , there 𝐷𝑖 is 𝑖-dimension, and 𝑛 = dim(𝐻) – dimensionality
of multidimensional
{︁ }︁cube [18]. Each dimension is characterized by a set of members
𝐷𝑖 = 𝑑𝑖1 , 𝑑𝑖2 , .., 𝑑𝑖𝑘 , there 𝑖 is a number of dimension, 𝑘𝑖 – the quantity of members.
𝑖
Members of 𝐷𝑖 are drawn from a set of positions of the basic classifier which corresponds
to an aspect of the observed phenomenon associated with 𝐷𝑖 .
The multidimensional data (︁cube is a structured
)︁ set of cells. Each cell 𝑐 is defined by
a combination of members 𝑐 = 𝑑1𝑖1 , 𝑑2𝑖2 , .., 𝑑𝑛
𝑖𝑛 . The combination includes one member
for each of the dimensions. If the analysis of the observed phenomenon is performed
using a large set of diverse aspects, not all members combinations define the possible
cells of multidimensional cube, i.e. the cells corresponding to a certain fact. This effect
occurs due to semantic inconsistencies of some members from different dimensions to
each other and generates sparseness in the cube.
The complex structure of the compatibility of members may lead to a situation
where a certain dimension becomes semantically uncertain if combined with a set of
members from other dimensions. In this situation, while describing the possible cell of
multidimensional cube the special value “Not in use” can be used to set the member of
semantically unspecified dimension. The structure of the multidimensional data cube in
the information system can be described as the set of possible members combinations.
Different values from the classifiers, which comply with the dimensions, and the special
Demidova Anastasia V., Fomin Maxim B., Shorokhov Sergey G. 67
value “Not in use” can be applied in the combinations of this set. To refer to the set of
possible members combinations we will use the abbreviation “SPMC”.
The subject domain is characterized by the measure values defined in possible cells
of the multidimensional cube. The full set of measures composes the set 𝑉 (𝐻) =
{𝑣1 , 𝑣2 , .., 𝑣𝑝 }, where 𝑣𝑗 is 𝑗-measure, 𝑝 – the quantity of measures in the hypercube.
Not all the measures from the 𝑉 (𝐻) can be defined in the possible cell. This situation
can appear in case of semantic inconsistency between the members defining the cell
and some measures. While describing multidimensional data cube structure for every
possible sell it is necessary to define its own set 𝑉 (𝑐) = {𝑣1 , 𝑣, .., 𝑣𝑝𝑐 }, which consists of
certain measures for this cell, 1 ≤ 𝑝𝑐 ≤ 𝑝. We can use the special value “Not in use” for
the description of c measures, which are not included in the set 𝑉 (𝑐).
The description of the SPMC can be obtained with the help of the cluster method
based on the analysis of links between members [19]. {︁ The cluster}︁ method allows
identifying the groups of members. The group 𝐺𝑖𝑗 = 𝑑𝑖1 , 𝑑𝑖2 , .., 𝑑𝑖𝑚𝑗 of members in
𝑖-dimension includes 𝑚𝑗 members (1 ≤ 𝑚𝑗 ≤ 𝑘𝑖 ), where 𝑗 is a group number and
contains members, which equally coincide in the SPMC with the members from some
groups of members of other dimensions.
It is possible to define connected groups of member in different dimensions with
the help of the semantic analysis. The cluster of members combinations 𝐾 is the set
of member combinations, which can be obtained with the help of Cartesian product
where operands are groups of members or special value “Not in use”; one operand stands
for every dimension used in the cluster SPMC(𝐾) = 𝐺1 × 𝐺2 × .. × 𝐺𝑛 . Clusters of
members combinations can be used for the description of the SPMC.
3. The use of classification schemes for describing the semantics of the
observed phenomenon
From the position of semantic the description of the observed pattern characteristics
within the multidimensional data model consists of the classification attributes detecting
(dimensions of the multidimensional cube) and establishing the links between them. It
can be rather difficult if there are a great number of dimensions. Classification schemes
(SC) can be used to solve the problem of classification of data describing the observed
phenomenon. For CS it is possible to formulate a number of requirements [20].
It is necessary to take into account the component structure of the observed pattern
while defining CS. If the observed pattern can be semantically divided into separate
structural components for which is possible to choose their own sets of aspects for
analysis every component should be compared with CS. The procedure of CS formation
is based on defining and analysis of the attributes relevant to the chosen aspects of the
analysis. The dimensions of the multidimensional cube should be compared with the
characteristics.
CS of the attributes for the observed patterns should be formed on the hierarchical
principle. Ranking should be established among the attributes related to CS. This ranking
allocates the dimensions which to some extent convey the essence of the structural
component for the observed pattern. This component is compared with CS. It is
necessary to define the major dimension which is more likely to reflect the semantic of
the structural component relevant to CS. The hierarchy of attributes should be formed
from other dimensions included into CS which are semantically subordinate to the major
dimension and express some particular properties of the structural component for the
observed pattern. The following principle should be observed: the members of the
major dimension convey the most important attributes of the observed pattern, the
members of other dimensions which come hierarchically below the major one convey
some subordinate attributes specifying the essence of the major dimension.
While forming the hierarchy of the attributes for the observed pattern in CS it should
be possible to describe the members of the major dimension separately or in groups of
members as different members can be connected with different semantic aspects of the
68 ITTMM-WSS—2018
structural component for the observed pattern. Different hierarchies of attributes should
be formed for the members of the major dimension which are semantically different.
In the hierarchy of the attributes in CS there must be the information about the set
of measures describing the observed pattern in case of choosing some particular members
from the hierarchy.
The classification scheme of the attributes for the observed pattern is an object of
the multidimensional information system which describes the structural component of
the observed pattern and contains the following information:
– the set of dimensions included into the classification scheme;
– the set of members of these dimensions included into the classification scheme;
– the major dimension chosen in the set of dimensions CS;
– the set of measures included into the classification scheme;
– the tree of member combinations CS which form the hierarchy of the attributes
included into CS.
The hierarchy principal of CS forming is realized in the structure of a tree which
presents the member combinations in CS. The tree of combinations can be formed as
a result of the semantic analysis of the structural component for the observed pattern.
The tree can be defined while describing the process of its formation. One should start
the formation of the tree from its roots where the groups of members of the major
dimension are placed. Then it is necessary to go down passing the hierarchical levels
and adding a group of members to each of them. Thus every group reveals the essence
of every previous member on the previous level. What is more it is necessary to add
the group related to the dimension which is mostly connected with the members of
the previous level. As a result different sequences of CS measures can appear in tree
brunches on the way from the roots to leaves.
Relationships between the members of different dimensions can be established us-
ing approaches of non-parametric methods of statistical analysis [21] and queueing
theory [22].
For the tree structure of classification scheme, the following rules must be followed:
1. The root of the tree is the unit “Major dimension”.
2. The tree itself is a hierarchical structure where the levels are set through alternating
such units as “Group of members” and units “Dimension”. At the same time groups
of members should be formed in the dimensions relevant to the units hierarchically
placed one level higher.
3. Leaves of the tree are units “Group of members”.
4. The unit “Group of members” (except the unit which is a tree leaf) should be
relevant to the unit “Dimension” hierarchically placed on a lower level. Only one
unit or several units “Group of members” placed on a lower level can be relevant to
the unit “Dimension”.
5. Moving from the root to a leaf you can see every dimension only once.
4. Semantic aspects of information system metadata construction
The analysis of the observed phenomenon reveals the qualification characteristics
that are included in the metadata of the information system. In the structure of a
multidimensional data cube, these characteristics are divided into subject of analysis
(measures) and aspect of analysis (dimensions). Semantic analysis allows to establish
connections between these characteristics. As a result, the structure of a multidimensional
cube can be revealed, that is, significant cells of a multidimensional cube corresponding
to the facts are described.
Difficulties in the application of the described technique are due to the fact that
the complex observed phenomenon is characterized by a large number of aspects. A
complete set of these aspects allows us to construct many multidimensional data cubes
corresponding to different structural components of the observed phenomenon. The
pairwise analysis of the characteristics does not make it possible to separate them
in accordance with the structural components, since they are mixed in the observed
phenomenon and there are no hierarchical relationships between them.
Demidova Anastasia V., Fomin Maxim B., Shorokhov Sergey G. 69
Construction of classification schemes related to the observed phenomenon as a result
of semantic analysis allows to achieve the following result:
– semantic separation of the characteristics of the observed phenomenon, their binding
to the structural components of the observed phenomenon;
– ranking of characteristics, building a hierarchy of characteristics in accordance with
their significance in the description of the properties of the structural component
of the observed phenomenon.
The described properties of classification schemes allow us to consider them as the
main objects that describe and systematize information about the structural components
of the observed phenomena. The interaction of information objects included in the
multidimensional data cube and the classification scheme can be represented by a
diagram (see figure 1).
Figure 1. Information objects describing the observed phenomenon
The links in the diagram represent semantic and technological dependencies. The
use of classification schemes alters the relationship between the observed phenomenon
and the multidimensional data cube. Classification schemes are a source of classification
of information objects of a multidimensional cube related to the structural components
of the observed phenomenon.
Members belonging to different dimensions and measures form a classification scheme.
The process of such formation is influenced by the semantic relationship between
the observed phenomenon on the one hand and the dimensions and measures on the
other hand. Due to the fact that the classification scheme expresses the properties of
the structural component of the observed phenomenon, there is an implicit semantic
relationship between the classification scheme and the observed phenomenon, which
forms the composition of dimensions, members and measures of the classification scheme.
The same relationship allows us to determine which classification schemes can underlie
multidimensional data cubes related to the observed phenomenon. The insertion of
classification schemes changes the nature of the relationship between the observed
phenomenon and the multidimensional data cube. The observed phenomenon loses its
role as a source of classification, while the multidimensional cube plays the role of a
technological component.
70 ITTMM-WSS—2018
5. Conclusions
The paper considers the method of designing information systems using a multidi-
mensional approach. This approach allows us to develop a system based on a metamodel,
which is semantically related to the subject domain of the system. The method is
based on the construction of groups of members which are semantically connected with
other groups of members. Connected groups related to different dimensions describe the
cluster of cells.
Classification schemes that correspond to the structural components of the observed
phenomenon can be used to describe it’s semantics. Every classification scheme is a
graph describing the hierarchy of members that are associated with a separate structural
component of the observed phenomenon. Classification schemes can be used to solve
the problem of classification of data describing the observed phenomenon. The coupling
between several classification schemes related to different structural components helps to
describe the metadata of the multidimensional information system. It is formed on the
hierarchical principle and establishes a ranking between the characteristics of structural
component of observed phenomenon. The classification scheme is a technological
component in relation to multidimensional data cube, while the multidimensional cube
plays the technological role in relation to the observed phenomenon.
Acknowledgments
The publication has been prepared with the support of the “RUDN University
Program 5–100”.
References
1. E. Thomsen, OLAP Solution: Building Multidimensional Information System, Willey
Publishing, 2002.
2. C. M. Hirata, J. C. Lima, Multidimensional cyclic graph approach: representing a
data cube without common sub-graphs. Inf. Sci. 181, pp. 2626–2655 (2011).
3. N. Karayannidis, T. Sellis, Y. Kouvara, CUBE File: A File Structure for Hierarchically
Clustered OLAP Cube, in: Advances in Database Technology, pp. 621–638, Springer-
Verlag (2004) ISBN 978-3-540-21200-3.
4. S. Chun, Partial Prefix Sum Method for Large Data Warehouses, in: Foundations of
Intelligent Systems — ISMIS 2003, pp. 473–477. Springer-Verlag (2004) ISBN 978-3-
540-39592-8.
5. R. B. Messaoud, O. Boussaid, S. L. Rabaseda, A Multiple Correspondence Analysis
to Organize Data Cube, in: Databases and Information Systems IV — DB&IS 2006,
pp. 133–146. IOS Press (2007) ISBN 978-1-58603-715-4.
6. R. Jin, J. K. Vaidyanathan, G. Yang, G. Agrawal, Communication and memory opti-
mal parallel data cube construction, IEEE Transactions on Parallel and Distributed
Systems. 16, pp. 1105–1119 (2005).
7. Z. W. Luo, T. W. Ling, C. H. Ang, S. Y. Lee, B. Cui, Range Top/Bottom 𝑘 Queries
in OLAP Sparse Data Cubes, in: Database and Expert Systems Applications —
DEXA’01, pp. 678–687. Springer-Verlag (2001) ISBN 978-3-540-42527-4.
8. L. Fu, Efficient Evaluation of Sparse Data Cubes, in: Advances in Web-Age Informa-
tion Management — WAIM’04, pp. 336–345. Springer-Verlag (2004) ISBN 978-3-540-
27772-9.
9. F. Z. Salmam, M. Fakir, R. Errattahi, Prediction in OLAP Data Cubes. Journal of
Information & Knowledge Management, 15, pp. 449–458 (2016).
10. O. Romero, T. B. Pedersen, R. Berlanga, V. Nebot, M. J. Aramburu, A. Simitsis, Us-
ing Semantic Web Technologies for Exploratory OLAP: A Survey, IEEE Transactions
on Knowledge & Data Engineering, 27, pp. 571–588 (2015).
11. L. I. Gomez, S. A. Gomez, A. Vaisman, A generic data model and query language
for spatiotemporal OLAP cube analysis, in: Proceedings of the 15-th International
Demidova Anastasia V., Fomin Maxim B., Shorokhov Sergey G. 71
Conference on Extending Database Technology — EDBT 2012, pp. 300–311. ACM,
New York (2012) ISBN 978-1-4503-0790-1.
12. M. F. Tsai, W. Chu, A Multidimensional Aggregation Object (MAO) Framework
for Computing Distributive Aggregations, in: Data Warehousing and Knowledge
Discovery — DaWaK 2003, pp. 45–54. Springer-Verlag (2003) ISBN 978-3-540-40807-
9.
13. J. S. Vitter, M. Wang, Approximate computation of multidimensional aggregates
of sparse data using wavelets, in: Proceedings of the 1999 International Conference
on Management of Data — SIGMOD’99, pp. 193–204. ACM, New York (1999),
ISBN 1-58113-084-8.
14. B. Leonhardi, B. Mitschang, R. Pulido, C. Sieb, M. Wurst, Augmenting OLAP
Exploration with Dynamic Advanced Analytics, in: Proceedings of the 13th Interna-
tional Conference on Extending Database Technology — EDBT 2010, pp. 687–692.
ACM (2010) ISBN 978-1-60558-945-9.
15. W. Wang, H. Lu, J. Feng, J. X. Yu, Condensed Cube: An Effective approach
to reducing data cube size, in: Proceedings of the 18th International Conference
on Data Engineering — ICDE’02, pp. 155–165. IEEE Computer Society (2002)
ISBN 0-7695-1531-2.
16. S. Goil, A. Choudhary, Design and implementation of a scalable parallel system
for multidimensional analysis and OLAP, in: Parallel and Distributed Processing —
11th IPPS/SPDP’99, pp. 576–581. Springer-Verlag (1999) ISBN 978-3-540-65831-3.
17. A. Cuzzocrea, OLAP Data Cube Compression Techniques: A Ten-Year-Long History,
in: Future Generation Information Technology — FGIT 2010, pp. 751–754. Springer-
Verlag (2010) ISBN 978-3-642-17568-8.
18. C. Chen, J. Feng, L. Xing, Computation of Sparse Data Cubes with Constraints,
in: Data Warehouse and Knowledge Discovery, pp. 14–23. Springer-Verlag (2003)
ISBN 978-3-540-40807-9.
19. M. B. Fomin, Cluster method of description of information system data model based
on multidimensional approach. In: Vishnevskiy V., Samouylov K., Kozyrev D. (eds.)
Distributed Computer and Communication Networks. DCCN 2016. CCIS, vol 678,
pp. 657–668, Springer (2016). doi:10.1007/978-3-319-51917-3_56.
20. M. B. Fomin, The application of classification schemes while describing meta-
data of the multidimensional information system based on the cluster method.
In: Vishnevskiy V., Kozyrev D. (eds.) Distributed Computer and Communica-
tion Networks. DCCN 2017. CCIS, vol 700, pp. 307–318, Springer (2017). doi:
10.1007/978-3-319-66836-9_26.
21. Y. Orlov, Y. Gaidamaka, E. Zaripova, Approach to estimation of performance mea-
sures for SIP server model with batch arrivals, in: Communications in Computer and
Information Science, vol 601 (2016) pp. 141–150. doi:10.1007/978-3-319-30843-2_
15.
22. Y. V. Gaidamaka, E. R. Zaripova, Session setup delay estimation methods for IMS-
based IPTV services. Lecture Notes in Computer Science (including subseries Lecture
Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol 8638, LNCS,
2014, pp. 408–418. doi:10.1007/978-3-319-10353-2_36.