EXTENDING THE UML-GEOFRAME DATA MODEL FOR

CONCEPTUAL MODELING OF NETWORK APPLICATIONS

Sergio Murilo Stempliuc, Jugurta Lisboa Filho, Marcus V. A. Andrade

Departamento de Informática – Universidade Federal de Viçosa (UFV)

36570-000 – Viçosa, MG, Brasil

Karla A. V. Borges

Prodabel – Empresa de Informática e Informação do Município de Belo Horizonte

Av. Pres. Carlos Luz, 1275, 31230-000 – Belo Horizonte, MG, Brasil

Keywords: Geographic databases, Conceptual data model, Network applications.

Abstract: This paper presents an extension of the UML-GeoFrame data model that includes a set of new constructors

to allow the definition of conceptual schemas for spatial database applications whose elements relationship

forms a network. Also, it is discussed how the GeoFrame conceptual framework is changed with the

inclusion of new metaclasses and the corresponding stereotypes related to network elements. The extension

proposed in this paper is evaluated using a class diagram for a water distribution company.

1 INTRODUCTION

Conceptual data models have been refined to meet

particular details found during the conceptual

modeling of geographical applications, allowing the

abstraction of geographical phenomena according to

spatial and temporal dimensions. These models also

allow the representation of geographical and non-

geographical entities, as well as the specification of

spatial, temporal and semantic integrity constrains.

Among several models created for Geographical

Information Systems (GIS), it is important to stand

out the balance between coverage and simplicity of

these models (Friis-christensen et al., 2001). As

pointed out by Kösters et al. (1997) and Spaccapietra

et al. (2006), some models define a large set of

elements trying to be general enough to be applied in

several domains, but usually, they are too complex

to be used by users and developers. Other models

are simpler and easier to use but are restricted and it

is necessary to include some additional textual

information to describe a specific domain.

A subject in GIS domain that has not been

receiving the necessary attention from authors of

these models is network applications, that is, those

applications in which the elements relationship

should be described by a network, such as water,

road, hydrographical, telecommunications and

energy distribution networks. One of the first

conceptual models trying to address this problem in

detail was the GeoOOA model (Kösters et al., 1997).

Other more recent models as OMT-G (Borges et al.,

2001) also provide simple constructors for network

modeling, but do not address the problem at the

same level of detail as the GeoOOA model.

The objective of this work is to extend the UML-

GeoFrame data model (Lisboa Filho and Iochpe,

2008) to provide constructors for the development of

class diagrams involving network elements. The new

constructors will allow the modeling of geographical

databases for domains that require mapping the

connectivity of their elements.

The article is organized as follows: Section 2

presents a general description of the GeoFrame

framework and the conceptual UML-GeoFrame data

model; Section 3 describes the proposed extension to

UML-GeoFrame for network modeling; Section 4

shows a class diagram of a water distribution

company; and Section 5 presents our conclusions.

164

Stempliuc S., Lisboa Filho J., Andrade M. and Borges K. (2009).

EXTENDING THE UML-GEOFRAME DATA MODEL FOR CONCEPTUAL MODELING OF NETWORK APPLICATIONS.

In Proceedings of the 11th International Conference on Enterprise Information Systems - Databases and Information Systems Integration, pages

164-170

DOI: 10.5220/0002002101640170

 SciTePress

Figure 1: The Geoframe Framework (Lisboa Filho and Iochpe, 2008).

2 UML-GEOFRAME:

A CONCEPTUAL DATA MODEL

FOR GIS APPLICATIONS

The UML-GeoFrame data model (Lisboa Filho and

Iochpe, 2008) proposes a set of stereotypes, forming

a UML profile (Unified Modeling Language) for

conceptual modeling of geographical databases.

Constructors of UML-GeoFrame data model are

based on the class hierarchy defined in the

GeoFrame framework that was originally proposed

in Lisboa Filho and Iochpe (1999). Section 2.1

presents an overview of GeoFrame where only the

elements needed to understand the new network

constructors are described. Section 2.2 shows the use

of the UML-GeoFrame stereotypes.

2.1 GeoFrame Overview

The GeoFrame framework (Figure 1) consists of a

hierarchical class structure that describes the

fundamental elements for the modeling of any

geographical database.

At the planning level, the geographic regions

corresponding to the areas of interest (Geogra-

ficRegion class) are defined and, for each region, it

is defined the associated themes (Theme class), such

as hydrograph, transport, relief, etc. A theme can be

subdivided in a hierarchy of sub-themes.

The metamodel level is composed of metaclasses

that reflect how the reality is interpreted and it can

be represented by conventional data (Conventional-

Object) or geographic phenomena (Geographic

Phenomenon). The latter is still specialized in

metaclasses for field (GeographicField) and objects

(GeographicObject) views.

Also, the level of spatial representation reflects

how the reality is represented by designers and users

more abstractly in relation to the representation

within the database. The SpatialObject class

generalizes spatial representation classes observed in

objects view, including Point, Line, Polygon and

ComplexSpatialObj. This last one represents a

spatial object formed by two or more spatial objects.

The FieldRepresentation class generalizes the field

view classes, including GridOfCells, AdjPolygons,

Isolines, GridOfPoints, TIN (Triangular Irregular

Network) and IrregularPoints. Multiple represent-

tations are supported in the two views.

2.2 UML-GeoFrame Overview

The UML’s package constructor corresponds to the

first concept incorporated in the UML-GeoFrame

data model. It is used at the planning level to

identify themes related to the geographic region,

allowing a top-down approach to the problem.

The UML-GeoFrame data model defines a set of

stereotypes which are important for the application

classes representation avoiding that these classes are

represented as subclasses of the metamodel level.

The aim is to avoid overloading the application

diagram and hindering its understanding. Figure 2

presents some classes examples with stereotypes of

ConventionalObject, GeographicObject and

GeographicField metaclasses, respectively.

EXTENDING THE UML-GEOFRAME DATA MODEL FOR CONCEPTUAL MODELING OF NETWORK

APPLICATIONS

165

Figure 2: Stereotypes for generalization/specialization.

Stereotypes are also defined at the spatial

representation level to represent different

abstractions of the spatial shape of geographic

phenomena. For example, in Figure 1, the

GeographicField class is associated to FieldRepre-

sentation class and GeographicObject associated to

SpatialObject. Based on these associations, it is

possible to model the spatial characteristic of a

geographic phenomenon. The stereotypes used for

the spatial representation were defined on these

associations (Figure 3). Then, two stereotypes are

used for each class, one reflecting the metaclass type

and another reflecting the spatial representation type.

A complete description of this model and examples

of use can be found in Lisboa Filho and Iochpe

(2008).

Figure 3: Stereotypes for spatial representation.

3 EXTENSION OF THE

UML-GEOFRAME MODEL

FOR NETWORK MODELING

This section describes the extension of the UML-

GeoFrame data model so that it will provide specific

constructors for modeling elements involved in a

network. Four stages of the GeoFrame framework

evolution are presented, identifying advantages and

disadvantages found in its previous versions. The

final version proposed is described in section 3.4.

3.1 Inclusion of Network Properties

into Geometric Classes

The first idea of GeoFrame extension was limited to

the inclusion of network properties into the spatial

Point and Line classes. The advantage of this

approach is to use the geometric characteristics of

point and line, considering the geometric

representation for the elements node and arc

respectively, although they do not depend on

information such as shape, length, position,

orientation and transformation. Another advantage is

the non-creation of new stereotypes to represent

these elements, only including attributes referring to

connectivity would be necessary.

Demirel (2002) mentioned the reduction in

requirements of data storage and increase in

performance of query processing as advantages of

this approach, but he pointed out the disadvantages

in relation to data maintenance. Although

topological data do not depend on geometric

properties, changes in object geometry would

necessarily cause the maintenance of both geometric

and topological data.

3.2 Specialization from Geometry

The second alternative for Geoframe extension was

to avoid the problems found in the first proposal

through specialization of Node and Arc classes from

Point and Line, respectively.

This solution has the same advantages as the

previous proposal, except for the inclusion of new

stereotypes, and solves the problem relative to

topological data maintenance when there is change

in geometric data. However, as similarly discussed

in Lisboa Filho and Iochpe (2008), a flaw exists in

this reasoning, as the generalization/specialization

concept does not recommend that different things

relate in such way.

3.3 Class Hierarchy for Network

Modeling

As the solution using the generalization/specia-

lization concept was shown inappropriate, it was

included a new hierarchical class structure to

provide the fundamental elements for the network

modeling. Figure 4 shows this new approach.

A new specialization was added from the

GeographicPhenomenon class at the metamodel

level, making it possible to interpret reality elements

as network elements through the NetworkObject

class. At the spatial representation level, the

NetworkRepresentation class generalizes the classes

of network representation, including Node, Arc and

ComplexNetworkObj. The last one represents a

network object formed by two or more network

objects.

Despite the hierarchical similarity with the field

and objects views, network modeling has some

particularities. The first refers to the multiplicity in

ICEIS 2009 - International Conference on Enterprise Information Systems

166

Figure 4: Class hierarchy for network modelling.

the association represent between the Network-

Object and NetworkRepresentation classes (Figure

4). As the association has multiplicity one to one

(1:1), an instance of the NetworkObject class should

be represented by only one instance of the

NetworkRepresentation class, i.e., there are no

multiple representations.

The second particularity refers to the association

between Node and Point, as well as Arc and Line, to

represent the geometric aspect of the network

classes. This characteristic imposes the restriction

that in any diagram using GeoFrame for network

modeling, there should be a class of the type Point

associated with a class of the type Node, in the same

way that a class of the type Line should be

associated to a class of the type Arc. These

relationships become optional when the application

involves only geographic phenomena in field and/or

object view.

The third characteristic refers to the Complex-

NetworkObj class. In spite of the apparent similarity

with the ComplexSpatialObj class, this class has a

different meaning. Each application of network

representation will always use the Complex-

NetworkObj class, since a network is an aggregation

of nodes and arcs. Being an aggregation, the same

node or arc may belong to several networks. The

ComplexSpatialObj class in its turn is not always

present in the modeling of an application using the

object view.

This solution, although meeting the basic requi-

rements for modeling of network applications, still

has some problems to be solved. The first problem

refers to the association between objects of

geometric and topological representation. As several

geometric characteristics are not pertinent to the

Node and Arc classes, their association with Point

and Line makes it possible to specify queries starting

from the topological classes with access to

characteristics that are not about connectivity. One

example is to query the attributes length and

thickness of a line associated with an arc.

Another problem refers to the network

representation through the ComplexNetworkObj

class. Since it is essential in network modeling and

does not add any spatial attributes to aggregated

objects, a new solution became necessary to replace

this class by a more abstract concept.

3.4 Final Proposal of GeoFrame

Extension for Network Modeling

To solve the problems of section 3.3, the GeoFrame

was changed as showed in Figure 5. The association

between Node and Point was removed, as well as

between Arc and Line, so that only the essential

geometric characteristics are used by network

elements. Besides, network elements can be

associated with different types of spatial objects. As

an example, both a power plant represented by a

polygon and a pole represented as a point can play

the role of nodes in an electric energy network.

However, the most important change was the

replacement of the ComplexNetworkObj class at the

spatial representation level by the Network class at

the metamodel level. As in the previous version

(Figure 4), only alphanumeric information

EXTENDING THE UML-GEOFRAME DATA MODEL FOR CONCEPTUAL MODELING OF NETWORK

APPLICATIONS

167

Figure 5: Final proposal of GeoFrame extension for network modelling.

was added to aggregated network objects, the

interpretation of a network became the same of a

conventional class, but maintaining the aggregation

of network objects. To make the interpretation of the

Network class clearer, as it was conventional class,

and show that each instance of the Network class

should belong to a package (Theme class), we

specialize the Network class from the

ConventionalObject class.

The aggregation between Network and

NetworkObject implies that in the application

diagram an instance of Network will be an

aggregation of NetworkObject instances, allowing

network handling as a composed object and allowing

that an instance of NetworkObject belongs to more

than one network.

A last modification was made in the hierarchy

proposal in order to specify bidirected and

unidirected networks. Although data structure for

nodes and arcs are different in these two types of

networks, the representation, on the other hand, only

differs by the arc shape. For the GeoFrame, the need

to specify the type of a network and to maintain

simplicity of model is based on representation,

therefore the idea consisted in specializing the Arc

class into the Bidirectional and Unidirectional

classes.

3.5 Network Stereotypes

With new classes added to the framework, new

stereotypes were also defined. They follow the same

UML-GeoFrame principles, with a stereotype

defined for the specialization of the geographic

phenomenon and another for spatial representation.

The Network class uses the conventional object

stereotype. Figure 6 shows three examples of classes

with the generalization stereotype for the

NetworkObject metaclass, and representation

stereotypes for the Node (a), Bidirectional arc (b)

and Unidirectional arc (c) metaclasses.

Figure 6: Stereotypes for network modelling.

3.6 Definition of Network

Representation Elements

Simplicity and expressivity are characteristics of the

UML-GeoFrame data model that make it easy to

understand the modeled reality. On the other hand,

they imply in additional textual information to build

a complete diagram concerning the problem domain.

UML-GeoFrame models, which provide simple

constructors for developing simple diagrams, also

have textual information to facilitate concept

understanding.

Textual information incorporated to models is

useful to define each of the classes or even to

establish restrictions common to several domains.

The Node class, e.g. is defined as a spatial

representation in network object view used when the

end or cross between lines is represented by a point

with connectivity. The Arc class is also a network

object used to express the connectivity between one

or two nodes through a line. Both are independent of

the real position or transformation applied to the

element of the portrayed reality.

ICEIS 2009 - International Conference on Enterprise Information Systems

168

4 A CASE STUDY

A conceptual data schema was designed for a Water

Distribution System of a city to demonstrate the use

of network constructors proposed. The example was

based on Barros et al. (1995) and was supported by

the staff of the Viçosa Municipal Water and Sewer

Service (SAAE, 2008). We present only the

fundamental elements of the system that was

restricted to consider only the part that conveys

water from the treatment plant to house

hydrometers.

Two networks of water distribution were

identified in a city: the main network, which

conveys water from the treatment plant to the

connection points with secondary pipelines, passing

through reservoirs; and the secondary, formed by the

distribution grid and consumer’s direct water supply.

The reservoirs provide water to supply the

demands of consumption, any emergency demands

and also, they keep the minimum or constant

pressure within the network. Regarding network

location, reservoirs can be located upstream of the

distribution network, for normal supply; or

downstream, to store unused water during hours of

lower demand and supply the network during hours

of higher consumption (Barros et al., 1995).

The water main stream is the largest-diameter

pipeline responsible for conveying water from the

treatment plant to the upstream reservoir. The main

pipelines are responsible for conveying water from

the reservoirs to the several connection points with

ramifications. This way, each stretch of the main

pipeline may have a reservoir and a connection point

or two connection points at its ends.

The main connection is the link between the two

networks, controlling the connection between a main

pipeline (larger diameter) and the network

ramifications (smaller diameter). Each stretch of the

ramification also has two more connection types,

one for connection between ramifications and

another one for connection between ramifications

and building ramifications, known as water-taking

device. The building ramification conveys water

from the public network to the consumer’s water

meter (hydrometer).

When nodes and arcs represent spatial objects, a

network is called spatially embedded. Examples of

this type of network include: road, electric power,

water and gas networks (Kösters et al., 1997). A

counter-example is the construction of a network

over adjacency relationships between

neighbourhoods of a city.

The concept of multiple inheritances in diagram

classes was chosen to elucidate the relationship

between the elements of a network and the

representation of spatial objects. To generate this

multiple inheritances cases, a class should be a

specialization of two order classes, for example,

GeographicObject and NetworkObject, and also, it

should associate the possible spatial representations

of both views. The replacement of these

specializations and associations are carried out by

using generalization and representation stereotypes.

It is still possible the use of multiple spatial

representations for objects view.

Figure 7 shows the UML-GeoFrame class

diagram designed for the water distribution system

in which most classes have four stereotypes. The

TreatmentPlant class, e.g., plays the role of node in

the main network with polygonal representation as

geographic object. The use of four spatial

stereotypes, in spite of simplifying diagram

visualization, did not avoid its overloading.

To prevent this situation, the UML-GeoFrame

has an important feature: the generalization

stereotypes can be left with no representation since

stereotypes of spatial representation for field, objects

and network views are disjoint groups. So, a class

can use only stereotypes of spatial representation.

Superclasses also assist designers and users since

they allow the identification of classes sharing

common features including generalization and

spatial representation.

The Figure 7 also shows that constructors for

aggregation and composition allow the definition of

the network(s) that the classes of type node and arc

belong to. Furthermore, the relationships between

arcs and nodes reduce the ambiguity during diagram

interpretation. The MainConnection class, for

example, aggregates to the Main and Secondary

networks, and relates to the Ramification and

MainPipelineStretch classes, connecting these two

networks.

Finally, Figure 7 also shows the use of

stereotypes for representation of unidirectional arcs.

Using this stereotype, it is possible to verify that

both networks in the example have a flow for water

distribution. However, it is not necessary the

presentation of some details such as initial and final

nodes. These issues are considered only in the

implementation phase, since the important thing

modeling this type of application is to determine the

type of network and its nodes and arcs.

EXTENDING THE UML-GEOFRAME DATA MODEL FOR CONCEPTUAL MODELING OF NETWORK

APPLICATIONS

169

Figure 7: Example of class diagram of a water distribution system.

5 CONCLUSIONS

This work presented the extension of the UML-

GeoFrame data model for conceptual modeling of

network elements. The proposed solution was based

on a set of specific requirements for these

applications identified by Kösters et al. (1997),

which include: (a) to show the incidence between

arcs and nodes; (b) to show which networks the arcs

and nodes belong to; (c) to make clear the class type

of the arcs, nodes and network without identifying

this types by class name; (d) to identify the classes

with common features (superclass); (e) to make clear

the class that represents the connection between two

or more networks.

Thus, comparing the solution proposed in this

article with the solution presented by the GeoOOA

model, it is possible to say that the UML-GeoFrame,

besides the good features listed above, it allows the

representation of specific associations that can occur

between elements of a network represented by nodes

and arcs. As well as, it facilitates the aggregation of

these elements with the composed object, which is

the own network.

To conclude, we believe that the conceptual data

schema can be read and interpreted with no

ambiguities even by non-specialist users and, at the

same time, it gives a larger freedom to specify

accurately the requirements of the application.

ACKNOWLEDGEMENTS

This project is partially financed by CAPES,

FAPEMIG and CNPq/MCT/CT-INFO

REFERENCES

Barros, Raphael Tobias de V. et al. 1995. Saneamento.

Manual de saneamento e proteção ambiental para os

municípios, UFMG. (In Portuguese).

Borges, K. A. V.; Davis Jr., C. A. and Laender, A. H. F.

2001. OMT-G: An object-oriented data model for

geographic applications. GeoInformatica, Vol. 5, No

3, pp. 221-260.

Demirel, H. 2002. An Integrated Approach to the

Conceptual Data Modeling of an Entire Highway

Agency Geographic Information System (GIS).

Library of Berlin Technical University.

Friis-christensen, A., Tryfona, N. and Jensen, C. S. 2001.

Requirements and Research Issues in Geographic Data

Modeling. In Proceedings of the 9th ACM

International Symposium on Advances in Geographic

Information Systems, pp. 2-8.

Kösters, G. et al. 1997. GIS-Application Development

with GeoOOA. International Journal of Geographical

Information Science, Vol. 11, No 4.

Lisboa Filho, J. and Iochpe, C. 2008. Modeling with a

UML Profile. In Encyclopedia of GIS, Eds. Shekhar,

S. and Xiong, H., Springer-Verlag, pp. 691-700.

Lisboa Filho, J. and Iochpe, C. 1999. Specifying Analysis

Patterns for Geographic Databases on the Basis of a

Conceptual Framework. In Proceedings of the 7th

ACM GIS, pp. 7-13.

OMG - Object Management Group 2006. Object

Constraint Language. Needham.

SAAE Viçosa - Serviço Autônomo de Água e Esgoto do

Município de Viçosa. 2008. (In Portuguese).

Spaccapietra, S.; Parent, C. and Zimányi, E. 2006. The

MADS Data Model - Concepts to Understand the

Structure of your Spatial and Temporal Data. In

International Workshop on Informative Modeling for

the Architectural Heritage.

ICEIS 2009 - International Conference on Enterprise Information Systems

170