Annales Univ. Sci. Budapest., Sect. Comp.

42 (2014) 261–282

APPLICATIONS OF HYPERGRAPHS IN
INFORMATICS:
A SURVEY AND OPPORTUNITIES FOR RESEARCH

Bálint Molnár (Budapest, Hungary)

Dedicated to Prof. András Benczúr

Communicated by Péter Racskó

(Received June 1, 2014; accepted July 1, 2014)

Abstract. After brief introduction about hypergraphs and their specific

capabilities that make them apt to be applied in various fields of research
within information systems, modeling and analytics. In the literature re-
view, we survey the fields that fit to hypergraphs theory and the applica-
tion of their capability to describe complex relationships. We outline our
research on modeling of information systems and architectures, and then
we propose a research agenda to exploit the rich representation ability of
hypergraphs in the context of various information and IT systems.

1. Introduction

Hypergraphs are able to describe complex systems as their descriptive power

is fairly strong because they are one of the most general graph and mathemat-
ical structures for representing relationships. The hypergraphs are the gener-
alization and extension of concept graphs and finite sets. The mathematical
Key words and phrases: information systems, hypergraph, modelling, enterprise architecture,
process algebra
2010 Mathematics Subject Classification: 68U35, 68M99, 05C65, 97M99, 68Q85
This work was partially supported by the European Union and the European Social Fund
through the project FuturICT.hu (grant no.: TAMOP-4.2.2.C-11/1/KONV-2012-0013).
We are very thankful to Prof. András Benczúr for raising the research issue and helpful
discussions.
262 B. Molnár

theory of hypergraphs was developed in the past decades; the generalizations of

definitions as trees, cycles, and coloring for hypergraphs have been elaborated
with the accompanying theorems.

The growing importance of computational science and informatics brought

in composite systems of large size. Several areas of modeling and structuring
of phenomena in cyberspace required more flexible representational approach
than traditional graphs. The system of complex relationships has made neces-
sary the further generalization of original definition of hypergraph to provide
a powerful descriptive method. Thereby the hypergraphs have found new ap-
plication areas and got a new impetus.

The reason is why hypergraphs seem apt to depict relations in informa-

tion systems, social networks, document centered information processing, Web
information systems, are the relationships among services within a service-
oriented architecture. One of the basic problems is that the concepts that po-
tentially may be mapped to vertices of graphs are heterogeneous. Consequently,
a theoretically and mathematically correct representation of the structure to be
dealt with cannot be achieved. For example, it seems to be a good idea to make
use of a bipartite graph for describing a system that contains two distinctive
set of things. Each single thing may be identified as a node of the graph, the
graph may contain two disjoint sets of nodes. However, this approach at the
same time loses homogeneity and heterogeneity feature of the system. The two
separate disjoint sets represent two distinct groups of things that are definitely
different but homogeneous within their own group. Although the operations
that required to handle components of the system may need a uniform ma-
nipulation of elements in some cases in spite of their dissimilarity from some
viewpoints.

The hypergraph and generalized hypergraph as concepts of discrete math-

ematics are appropriate for describing such systems. The basic differences at
conceptual level between the traditional graph and hypergraph theory is that
a specific edge tow nodes in a graph, however, in a hypergraph - the so called
hyperedges - can group together more than two vertices. The generalized hy-
pergraph yields the opportunity that a hyperedge can be perceived as a vertex
and be a component of other hyperedges as a node. Consequently, the hyper-
graph is a generalization of several mathematical concepts as graphs, projective
planes, affine planes [8].

Firstly, we will survey the application areas of hypergraphs in informatics

and information systems, and then we will present our research in the context
of Web information systems and their architecture. Finally, we will outline a
Applications of hypergraphs in informatics 263

research agenda in the domain of information systems for further application

opportunities.

The structure of our paper is organized as follows. In Section 2, we outline

applications of hypergraphs for social networks analysis. This is followed by
Section 3 that describes hypergraph applications in a SOA and Cloud environ-
ment. We provide in Section 4 a description of our research on modeling and
perceiving the architecture of information systems. Finally, conclusions and
future work directions will be shown in the last section.

2. Social networks analysis

We have studied earlier the alternatives of software and system architec-

tures that fit to the investigation of social networks [23]. There are studies on
structure and relationships within social networks as e.g. Twitter [29]. Within
the representation of On-line Social Networks (OSN) or Social Network Sites
(SNS) as graphs, the vertices designate individuals or group of persons. The
links among them represented as the edges of graphs and may describe friend-
ship (with various granularity), cooperation, business relationships etc.

However, Twitter as social network has a fairly complex construction, namely

it has concepts of relationship as following, reply to and mention. That leads
to a more complex graph representation that requires directed, differentiated
edges between vertices that can be represented e.g. by coloring. The interde-
pendencies between the various messages should be taken into account if any
semantically rich interpretation is required for the analysis, namely either the
topic of texts or the relationship among the messages or posts should be shown
in an appropriate representational way [2, 31].

Facebook basically and originally contained a friendship relation. On evolv-

ing of services at Facebook, the relationship of friendship has been refined
as close friends, friends, and friends except acquaintances, public. These at-
tributes are used in sharing micro-blogs and posts. The basic relation between
people and groups became a more sophisticated system. There are solutions to
represent such structure by graphs but hypergraphs offer a uniform treatment
of the above mentioned phenomena.
264 B. Molnár

One of the specificity and area of interests of Twitter is that the Twitter
data are not only a set of data items and texts but through the contained
data items a complex network can be built up. One of the structure is the
follower graph (tweeters that follow specific other users), the other relation-
ship that is fairly loose and ad hoc - is created by the mechanisms of re-tweet
and hash-tag. The previously mentioned network structures can be mapped to
graph structures but to discover the basic organizing principles of sub-graphs
leads to challenging graph theoretic questions and empirical investigations. The
various features of relationships between individual entities can be reflected
better by hypergraph than traditional graph structure. The links displaying
multi-faceted characteristics in the case of Twitter and Facebook as well can be
mapped by hyperedges more faithfully than other graph theoretical approaches.

A research question can be formulated whether the relationship graphs of

two deeply different SNSs as e.g. Twitter and Facebook provide some clues
about the evolution of networks, features of users behavior that can be spotted
by graph structures.

OSN frequently contains so called supra-dyadic relationships [6], i.e. the

dyadic relationships appear between pairs of actors. On investigation of tweets
within Twitter, the attributes of tweets that are worth analyzing not only the
actors playing a role in actual tweets but e.g. geographical positions of actors
through geo-tagging furthermore the applied language, date and time of tweets
etc. In database worlds or entity-relationship modeling, we may call these
types of relationship as ternary or quaternary relations among entities. When
complex social networks are represented by hypergraphs the supra-dyadic rela-
tionships can be examined through the eigenvector centrality of hypergraph [6].
The vertices may represent individuals that produce tweets, the nodes linked
by hyperedges manifesting follower, re-tweet, hash tag relationships extended
by geographical, time and other necessary information.

3. Service-oriented architecture

Service-oriented architecture (SOA) turns into a popular software architec-

ture as enterprise architecture that supports the business processes at concep-
tual, system analysis and logical design viewpoint. Service-oriented architec-
ture provides a toolset to assist in application development and enterprise wide
application integration. The essential characteristic of SOA is, that the compu-
Applications of hypergraphs in informatics 265

tational resources are accessible in the form of services [15]. The past decade
has brought success for this architectural approach at several companies be-
cause this solution is considered as cost-effective and one that takes care about
reusability.

The most recent development on architecture area led to the concept of

SaaS (Software as a Service) within a Cloud environment. The tenants of
Cloud use various services in the form of SaaS applications that are built up
from components out of SOA services. In a multi-tenant environment, the
services are shared among several customers. The multi-tenant applications
yield benefits both for customer and service provider since the multi-tenant
solution decreases service delivery costs for service provider and reduces the
subscription price for the customer or enterprise. Although, the multi tenant
environment leads to issues to be solved. Namely, the Quality of Service (QoS)
is a critical point how to assure a service level when multiple tenants share the
same set of applications [30]. The performance issues should be handled as e.g.
(virtual) servers housing specific services, may be overloaded by the tenants;
the service level as e.g. response time, transaction output, and availability and
as a consequence of saturation reliability for these services will be degraded.
The service levels described in QoS should be mapped onto services consisting
of applications for each single tenant. The load balancing of (virtual) servers
is vital to keep the service levels at the pre-defined values in order to fulfill the
QoS for each single tenant. It is supposed that public clouds has a practically
infinite capacity and elasticity [32]. In the case of private and/or community
cloud, the statement mentioned before is not necessarily valid therefore a model
for load balancing and sharing of services is needed.

The services in a cloud environment can be grouped into three classes

namely: business-independent, business-dependent and compound business ser-
vices. The types of mutual dependencies between services from the cloud could
be functional, business and platform. The dependency appears at logical and
physical level that primarily relates to technology architecture layer that is typ-
ically hidden before tenants. The services can be mapped bijectively on vertices
of hypergraphs whereby the vertices in the hypergraph can be placed into par-
titions of the hypergraph by their mutual dependencies, by their deployment
on service centers, and the requirement of algorithms dedicated to load balanc-
ing [12]. Elastic Service Placement Problem (ESPP) applies a combinatorial
auction problem that may be formulated in a hypergraph terminology as well
[7]. Hypergraph partitioning is a very beneficial approach for load balancing
purposes in the case if data about dependencies and connectivity are obtain-
able. The general opinion is that hypergraph partitioning provides a better
and superior method than the traditional graph partitioning in the case of load
266 B. Molnár

balancing [5]. There are concurrent approaches in the literature [11]. The hy-
pergraph model fits to concurrent, communicating processes e.g. services in a
SOA for business processes. One approach is that the data or document ob-
jects can be designated as vertices and then the communication demands can
be specified by hyperedges. The hypergraph can be represented by matrices
by which the communication cost and volume very precisely can be rendered.

The hypergraph model and partitioning are well-suited to accurate descrip-

tion of communicating processes, their resource and capacity requirements, and
the sub-optimal but efficient algorithms can be designed to solve practical prob-
lems.

4. Systems modelling

As the literature proves it, the graphs as a representational structure are

apt to system modeling, analysis and specification, especially to information
and software systems [10, 14]. One of the opportunities is the use of the com-
bination of graph and algebraic means [16]. In spite of their expressive power
and flexibility the challenge is to find an appropriate representation of systems
being investigated. The representation should accurately describe the states
of systems, i.e. system configurations, keeping up the isomorphism between
the structure of systems and representations through the graphs. The isomor-
phic mapping between systems and graph representation has several rewards.
The graph theoretic approach yields a formalism that provides uniformity in
some sense and a normal form representation. The mathematically sound
representation makes the reusability possible. Moreover, the exploitation of
mathematical tools for system analysis and formal consistency checking are
available. The methods that can be applied are as follows: graph transforma-
tions, matrix representations of graphs, furthermore computation of properties
and feature based on mathematical theorems. These methods can assist to
assess soundness of model represented by graphs. For example, the correctness
of mapping can be checked by test algorithms for isomorphism. The validity of
transliteration of systems into graph representation is critical for exploiting the
available mathematical tool set to model both the dynamical and static facet
of systems as e.g. operational semantics, re-configurations, controls, security
and the model transformation in analysis, design and operation time.

The representation of systems’ characteristics as e.g. states, configurations

Applications of hypergraphs in informatics 267

through algebraic methods and techniques can be transcribed into graph repre-
sentations. The typical algebraic method can be one of the versions of process
calculi. The system representation appears in algebraic terms. The states and
system configurations formulated in process algebraic syntax can be mapped
into graph representation through inductive procedures. For Web Information
Systems, a combined method was developed that contained process algebraic
descriptive tools and methods of conceptual modeling that are especially suit-
able for designing information systems operating on the Web. The story algebra
provides a rich tool set that can capture the sides of processes and the use-cases
[27]. The processes within information system are derivatives of business pro-
cesses; the actual use of them is described by scenarios and scenes. The story
algebra offers the opportunity to treat with the Business process model and
workflow side and the end-user/ organization side of Information System Mod-
els [4]. Moreover, the story algebra makes possible to link the process and
end-user perspectives with the documents that are transporters of data and
information, the structure of documents has some correlation with the struc-
ture of organization. Conway’s thesis claims the correspondence between the
structure of development team and produced software architecture [19]. The
correlation between the document structure linked to an Information System
and the structure of the organization can be assumed.

5. Modeling document intensive web information systems

The most recent information systems can be regarded as Web Information

Systems (WIS) as Web became ubiquitous and a lingua franca for man-machine
dialogues through various protocols that are used in the context of end-users.
The WIS decisively differs from a set of Web pages. The differentia specifica
between collection of Web pages and WIS can be perceived by the following
way; WIS supports business processes and typically strongly integrated to other
information systems within organizations.

The Service Oriented Computing (SOC), Cloud Computing creates a new

context in which the information systems modeling should be re-thought [21].
The other important aspect of changes is the emphasis on documents as medium
for information exchange, the documents can appear in the forma of HTML
pages, SOAP messages, semi-structured documents (XML), and unstructured
documents [3, 13, 26].

The information systems modeling is a challenging task, for this reason sev-
268 B. Molnár

eral modeling frameworks came into existence. To understand in a holistic way

and a top-down approach of information systems, the Enterprise Architecture
frameworks offered a comprehensive view as Zachman and TOGAF method [18,
33]. Blokdijks Four Models in Three Views of Information Systems presents
principles for analyzing and designing information systems. The information
systems are inherently complex systems so that the modeling efforts should keep
an eye on following principles: graphical modeling languages provide the best
human interpretable description of applications; the model instances should
be transformed into the refined model instances. The documents in the form
of XML should join some of the model instances that represent significant
business processes. The combination of Zachman Enterprise Architecture, the
axiomatic design, document object model (DOM) and the use of UML as vi-
sual language form an approach that can be used for modeling WIS [22, 24, 25].

5.1. A formal hypergraph model

Web Information System (WIS) We create an abstraction of WIS in the

form of a generalized hypergaph that
consists of vertices and hyperedges.
Node/vertex in a hypergaph Each node (or vertex) corresponds to an
element within a WIS, e.g. documents,
elements of documents (constituting a
tree structure), business processes, pro-
cesses in scenarios and scenes, work-
flows, layers of workflows, web services,
networks of web services, etc.
Edge in a hypergaph Edge is a specific hyperedge with car-
dinality equal to two. Edge denotes bi-
nary relationships between two nodes,
as e.g. free documents is consumed by
a certain Web service, a generic docu-
ment is the ancestor of an intensional
documents, a free-document resulted in
a ground-document after binding, valu-
ating of variables etc.
Applications of hypergraphs in informatics 269

Hyperedge A hyperedge represents a relationship

among a subset of nodes as e.g. Web
services belonging to a specific work-
flow, Business Process containing work-
flows, nodes of a document tree struc-
ture composing a document etc.
System graph A hypergaph that includes a disjoint
node dedicated to the modeling envi-
ronment of the system, plus all the
nodes and hyperedges of the WIS.
Sub-system A subset of nodes and their incident hy-
peredges. A node/vertex is incident to
a hyperedge if the hyperedge contains
the node/vertex. A sub-system may be
composed of documents, Web services
and related entities out of data model
etc.
Subsystem may be created by opera-
tions that conforms with graph theoret-
ical concepts as:
Induced subhypergraph;
Subhypergraph;
Partial hypergraph;
Interconnecting sub-systems - A graph consisting of all the vertices
hyperedges graph of the general- in sub-systems and all hyperedges con-
ized hypergaph necting subsystems along with the dis-
joint, environment node.

Table 1: Mapping the Concepts of Web Information Systems onto the Notion
of Hypergaph.

Firstly, we should bring to mind the basic definitions of hypergraphs in or-

der to apply for modeling of WIS.

Definition 5.1 (Hypergraph). A hypergraph H is a pair (V, E) of a finite

set V = v1, .. ., vn and a set E of nonempty subsets of V. The elements of V
are called vertices, the elements of E edges [8].

Definition 5.2 (Generalized hypergraph). The notion of hypergraph may

be extended in a way that the hyperedges can be represented in certain cases as
vertices, i.e. a hyperedge e may consist of both vertices and hyperedges as well.
270 B. Molnár

The hyperedges that are contained within the hyperedge e should be different
from e.

5.2. Hypergraphs and information system modeling

The proposal that we want to present is to use hypergraphs in the Web

Information Systems modelling in various viewpoint and perspectives of In-
formation Systems Architecture. We hope that the hypergraphs can provide
a unified and uniform framework to handle the complexities and diversity of
models that are employed during the analysis, design and operation of informa-
tion systems. From theoretical viewpoints, the hypergraphs may provide better
understanding of organization of information systems and structuring princi-
ple. From practical viewpoints, it gives clues for designing and operational
principles and takes care of controlling and security mechanisms.

5.3. Web information system architecture

Generally, the Information System Architecture describes the structure of

modules of information processing systems, the links among components, the
design and analysis principles at information system level of which the main
purpose is to buttress business processes.

The logical design of information systems in the sense of technology appears

in the form of software architecture and their building blocks. The Zachman
Framework [33] was one of the first attempts that stressed the fact that soft-
ware architectures were not enough to capture the essential characteristics of
Information Systems Architecture. While software architectures describe de-
tails of systems structure (using, for example, E-R and DFD diagrams, Gang
of Four architecture pattern), Information System Architecture focuses on the
high-level business and information system processes [17]. Enterprise IT Ar-
chitecture can be understood as the collection of strategic and architectural
principles that includes the Information, Business System, and Technical Ar-
chitectures:

• Business (systems) architecture. The major constituents are the business

processes, the related roles and actors, furthermore, the structure and
information content and functional services of all business systems within
the organization.

• Information (and Data) Architecture renders the data types, information

Applications of hypergraphs in informatics 271

and containing documents that assist business processes; moreover the

interrelationships among pieces of information.

• Technical Architecture contains the software and hardware technologies

in the form of logical and physical platforms used in application im-
plementation in eh form of information systems. Technical architecture
sustains the integrity of the hardware, software, and infrastructure envi-
ronment that is necessary to support the Business Systems Architecture
and Information Systems Architecture.

Each single architecture layer can be refined from the viewpoints of stakehold-
ers. The Zachman framework demonstrates a solution for a higher granularity
of viewpoints and views that is able to express more accurately the complex
relationships between the models of information systems.

Up to now, one of the best systematic structure of information models is

given by Blokdijks work [4]. The main parts of Blokdijks structuring framework
are: organizational model that represents hierarchy of business units and ele-
ments and the method of work; information model that contains information,
structured, semi-structured, unstructured and other media format material,
furthermore their origin and course of action for manipulation; data model
that represents things of the physical world about which information is stored
along with their links to each other , it lays the ground for the implementation
model of data; process model that yields a description about the structure of
business activities and the strongly coupled to the control structure.

The Zachman framework and the Blokdijks information systems model are
in fact two orthogonal views of information systems. Thereby, a three dimen-
sional structure came into existence. Some components of Blokdijks informa-
tion systems model are used in specific viewpoints and perspectives of Zachman
Enterprise architecture. We can arrange that complex structure in a cube. The
interrelationships can be captured by the generalized hypergraph approach.

The Blokdijks model differentiates between data and information model

that fact provides the opportunity to handle separately the documents as in-
formation resources. A document in itself is a tree structure that can be de-
scribed by the document object model along with XML [22]. The structure of
the document can be represented by hyperedges, but at the same time, the set
of hyperedges manifests the document can be perceived as a vertex that belong
to other hyperedges that demonstrate the use of the document in other parts
of the model. For example, the document is used in business process within
workflows, the document may play the role of input and output for use cases in
B. Molnár

J. A. Zachman S. Entities = what Activities= how Locations= People = who Time= when Motivation =why
H. Spewak Data Architec- Applications whereTechnology
ture Architecture Architecture
Planner Objec- List of Business List of Business List of Business List of Organi- List of Events List of Bus. Scope
tives / Scope Objects Class = Processes Process Locations Node zations important Significant to Goals / Strate-
(Contextual) Class of Business = Class of Busi- = Major Business to the Business the Bus. Time gies Ends /
Thing ness Process Location People = Major = Major Bus. Means = Maj.
Organizations Events Bus. Goals /
Crit, Suc. Factor
Owner Enter- Semantic Model Business Process Business Lo- Work Flow Model Master Schedule Business Plan Enterprise Model
prise Model Object Class = Model Proc. = gistics System People = Organi- Time = Bus. Ends = Bus.
(Conceptual) Business Entity Bus. Process, Node = Busin. zation Unit Work Event Cycle = Objective Means
Association = Web Services Location Link = = Work Product, Bus. Cycle = Bus. Strategy
Bus. Relation- I/O = Bus. Business Linkage Documents
ship Resources, Docu-
ments
Designer Infor- Logical Data Application System Geo- Human Interface Processing Struc- Business Rules System Model
mation Systems Model Ent. = Architecture graphic De- Architecture Peo- ture Time = Sys- Ends = Struc-
Model (Logical) Data Entity Proc. = Appli- ployment Ar- ple = Role Work tem Event, Or- tural Assertion,
Reln = Data cation Function, chitecture e.g. = Deliverable, chestration Cycle Means = Action
Relationship Web Services, Distributed Sys- Semi-structured = Processing Cy- Assertion,
Method of Ob- tem Arch. Node documents cle
ject Class, I/O = I/S Service.
= User Views, (Processor, Stor-
Semi-structured age, Logical
documents Application Com-
ponent. etc.)
Link = Rela-
tionship between
Logical Appl.
Comp.
Builder Technol- Physical Data System Design System Architec- Presentation Ar- Control Struc- Rule Design Ends Technical Model
ogy Model (Phys- Model Ent. = Proc. = I/S Ser- ture / Technol- chitecture People ture Time = = Condition
ical) Segment / Table vices I/O = Data ogy Architecture = Screen Format, Execute, Chore- Means = Action
/ etc Reln = Elements / Sets, Physical Applica- HTML / XML in- ography Cycle =
Pointer/ Key / XML / HTML tion Comp. Node terface Work = Component Cycle
etc documents = Hardware / User
Systems Software
Link = Line
Specifications
Subcon-tractor Data Definition Programs Sup- Network Archi- Security Archi- Timing Definition Rule Specifi- Components
Detailed Specifi- Repository Ent. porting Software tecture Node = tecture People Time = Interrupt cation Ends =
cations (Out-of- = Field Reln = Components Address Link = = Identity, Au- Cycle = Machine Sub-condition
context) Address Proc. = Lan- Protocol thentication, Cycle Means = Step
guage Statement Authorization,
I/O = Data Item, Work = Job
XML Field
Functioning En- Data Function Network Organization Schedule Strategy
terprise
Table 2: A schematic mapping between Zachman’s architecture and the WIS component
272
Applications of hypergraphs in informatics 273

the sense of UML. The use cases can be formalized in scenarios and scenes for
WIS. The scenarios contain processes that are formulated by process algebra.
The hypergraph is able to display the before-mentioned complex relationship.

5.4. Hypergraphs and systems engineering

The hypergraph approach has the advantage of reducing the degree of com-
plexity and variety of relationships between models thereby the WIS can adapt
to the changing environment better, and enhance the stability of system.
The hypergraph approach can be applied to the information and data model.
As we have discussed earlier, the hypergraph is capable to represent even the
complex document structure. The document object model (DOM) can be
rewritten in hypergraph structure and this way the document-centric feature
of WIS can be represented by a better way, the structure of interrelationships
among process, organization, core data structure and document elements can
be expressed [9]. The evolution of documents and the related data can be
tracked more easily both in design and operation time [24]. The clash be-
tween the inherent nature of documents and disciplined data structure (e.g.
database) can be resolved through hypergraphs. The data structure, for exam-
ple, in relational database format can not be divided into parts - i.e. partitions
of databases - that correlates to the organization structure; nevertheless the
document structure follow more or less the structure of organization and the
roles within it. Utilizing the hypergraph approach, the dual nature and be-
haviour of documents can be handled, i.e. the structure of documents aligned
with the organization and the data structure being a definite part of database
and contained in documents can be reconciled into a unified framework.
The design artifacts of WIS can be perceived as the set of certain documents
and the strongly coupled set of processes incorporated in the form of scenarios
and scenes in the sense of story algebra approach [27].

The generalized hypergraph approach makes possible to build-up structures

that are orthogonal to each other although the complex of structures can ex-
press the systems of relationships in a disciplined way. The information space
consists of documents, data contained in documents, and the information bases
manifesting in traditional structured and semi-structured databases.

The documents as input and output media and the elements of story alge-
bra compose together the design artifacts of WIS. The basic concepts of story
algebra can be represented in a hypergraph by the following way:

• Story space, is the information space of WIS. The elements of information

274 B. Molnár

space as e.g. documents and group of relevant data appear as nodes of

hypergraph.

• Story board, can be perceived as directed, generalized hypergraph. Hy-

perarc (directed hyperedge) represents the evolutionary development be-
tween documents or document types. The documents or document types
are represented by by document vertices that are indeed hyperedges. The
hyperedges reflect the internal structure of the documents.

• Scenario, is a subdirhypergraph (directed subhypergraph) of the story

board that describes the transformation of inputs in the form of docu-
ments and data through steps of processes.

• Story, is a directed path within the story board. The story board is a
directed hypergraph (dirhypergraph). The starting point of the path is
an instance of document type that contains a collection of free variables;
the end-point is a document that contains all variables in fully evaluated
form, moreover linked to the underlying databases.

• A scene can be represented as hyperedges and hyperarcs. Some of hyper-

arcs consist of documents and document types that can play the role of
tail and head of the hyperarc and at the same time designate the changes
of documents. Other incident hyperarcs contains as head a role (job re-
sponsibility) and/or actors of interested organization units so that the
hyperarc designates the responsibility for manipulation on the specific
document by roles and belonging actors; and as tail the role or actor
who is responsible for receiving the document. From the process-oriented
view, a scene is elementary part or task of information systems services,
or a kind of Web services;

• The life cycle of documents can be described by a directed path through

the dirhypergraph, the documents may be contained one or more sce-
narios. The documents without free variables appear as head of certain
hyperarcs as the final state of documents manipulation.

v2 v3 v4 v5 v7 v8 v10 v11
E1 1 1 0 0 1 1 1 0
E2 1 0 1 1 0 0 0 0
E3 0 0 1 1 0 0 1 1

Table 3: The incidence matrix representing the hyper network

Applications of hypergraphs in informatics 275

Figure 1: A part of a directed hypergraph for story board

5.5. Model transformations

To put the disparate and different models into a unified framework, the
first attempt is to map the WIS and its possible models into the Zachman
architecture that contains the important aspects and views of an IS within an
organization context. To make the various models comparable, verifiable in
the sense of correctness, the object-oriented modeling approach can be used
for each modeling artifact as all the views of Zachman can be modeled by
object-oriented paradigm from the classical IS and database technology to the
document-oriented WIS.

Beside the representation of document and data, the description of appli-

cation in the form of WIS could be done in an object-oriented manner. The
UML provides a visual language, at the syntactic level, the systems analysis
and design methods offer the systematic way how to organize a system into an
operational application [20].

Axiomatic Design (AD) Theory [28] is a design method that offer assis-
tance for designers to structure design problems by a systematic way. The
axiomatic design approach can be applied to modeling and design of WIS. The
276 B. Molnár

Figure 2: Partitions of the hypergraph according to document-centric story

algebra

story algebra, the processes and documents as design artifacts appears at the
surface of WIS. The processes captured through the scenes and scenarios of
story algebra should be further refined to meet the underlying information and
data bases. The consistency and integrity between the document-centric story
algebra representation and the internal part of WIS that consists of the data
structure, the behavior of entities and the impacts triggered by events.

The axiomatic design can be aligned with the Zachman architecture as

perspectives can be considered as domains as all domains of axiomatic design
theory is contained within the set of Zachmans perspectives and represents the
life cycle of system development and operation.

The models that appear in the columns of Zachman architecture can be

created using of the object-oriented paradigm. The framework matrix consists
of a vertical axis that provides multiple perspectives of the overall architecture
and a horizontal axis, which provides a viewpoint of the stakeholder. Each
single model appears as a specific design artifact within the overall architecture.
The perspectives and classifications within the framework are illustrated in
Table 2.
According to axiomatic design theory, the requirements can be captured by the
Applications of hypergraphs in informatics 277

following concepts: Customer Needs (CNs), Functional Requirements (FRs),

Design Parameters (DPs), and Process Variables (PVs).

Figure 3: Hyperarcs expressing roles and responsibilities, documents evolution

and interdependencies between them

Combining the hypergraph approach and axiomatic design methods yields

a robust design tool. The hyperedges of the hypergraph can show us which
DPs satisfy which FRs, which model refines which models, which models fulfill
certain DPs or FRs.
The analysis and design of WIS consists of refinement and process of decom-
position that proceed through the matrix both vertically and horizontally. The
mapping process between the perspectives and viewpoints within the Zachman
matrix can be expressed mathematically in terms of the characteristic vectors
that define the design goals and design solutions, furthermore the relationships
and their constraint. At a given level of design hierarchy, the set of functional
requirements that define the specific design goals constitute a vector {FRs}
in the functional aspect. Similarly, the set of design parameters in one of the
perspectives for the FRs also constitutes a vector {DPs}. The relationship
between these two vectors can be written as

(5.1) {F Rs} = A{DP s}.

The matrix A type represents the actual method, the transformation and
mapping between the functional requirement and design decision (5.1). The
278 B. Molnár

verification and validation for properties of correctness, faithfulness, consis-

tency and integrity among the design decision is represented by matrix type B
(5.3). Matrix B designates the state transition between the relevant pairs of
models situated in the columns, i.e. aspects of Zachman’s architecture Table
4.

 
a11 a12 a13
(5.2) [A] = a21 a22 a23 
a31 a32 a33

The User Requirements, the Customer Needs (CNs) transformed into Func-
tional Requirements (FRs). The Functional Requirements can be represented
by models that refers to classes of objects in Table 4.

(5.3) {DP s} = B{DP s}

Matrix B representents the permanent refinement of the design that defines

the characteristics of the process and is similar in form to A. The naviga-
tion rules for transformation can be deduced from the incidence and adjacency
matrix of the hypergraph that represents WIS. The design hierarchies, the
perspectives of Zachmans architecture are described in a unified and uniform
manner by the hypergraph.

Q Aspects
Q
Q Entities Activities Locations People = Time = Motivation
Q = what = how = where who when = why
Perspectives Q Data Appli- Tech-
Q Archi- cations nology
tecture Archi- Archi-
tecture tecture
Contextual CN CN CN CN CN CN Scope
Conceptual FR FR FR FR FR FR Enterprise
Model
Logical DP DP DP DP DP DP System
Model
Physical DP DP/ DP DP DP DP Technical
Model
Detailed repre- PV PV PV PV PV PV Compo-
sentation (out- nents
of-context)
Functioning Data Function Network Organiza- Schedule Strategy
enterprise/orga- tion
nization

Table 4: Model transformation based on Zachman architecture and CNs, FRs,

DPs, PVs
Applications of hypergraphs in informatics 279

6. Conclusion

The mathematical formalism has put the disparate approaches as Blokdikjs

Information Systems Model, Zachman architecture/ontology and the Axiomatic
Design into a unified framework. The hypergraph approach provides a tool for
integrated management of that the orthogonal viewpoints of an information
system as data and information, the procedural aspect representing the system
behavior, and the models as design artifacts. The Axiomatic Design offers the
systematic and mathematical formalization of requirements and their refine-
ment during the system life cycle. This environment helps to change manage-
ment of both design and operational time of WIS. The hypergraph accurately
represents the interdependencies and set of relationships among the elements
of WIS. These relationships are mathematically described by the adjacency
and incidence matrices. On one side, the formal matrix descriptions can assist
in checking the consistency, integrity, accuracy of WIS. However, the changes
cannot be avoided either in design or operational time. The changes in the
design time should be escalated through the models, the design and engineer-
ing activities. The propagation of modification will criss-cross the aspects and
perspectives of Zachman matrix, between the various meta models for handling
certain single models. The hypergraph approaches take care of consistency of
models.

The most modern databases provides opportunities to operationalize our

ideas. The formalized mathematical description can be implemented by graph
databases. Moreover, several hypergraph databases emerged in the open source
domain that offer services for implementing hypergraph structures directly in
this database environment [1]. As further research, a prototype implementation
of the formalized approach will be attempted.

References

[1] Hypergraphd. http://www.kobrix.com/hgdb.jsp, 2014.

[2] Abraham, A., A.-E. Hassanien and V. Snasel, Computational Social
Network Analysis: Trends, Tools and Research Advances, Springer, 2009.
[3] Bernauer, M. and M. Schrefl, Self-maintaining web pages: from theory
to practice, Data & Knowledge Engineering, 48 (1) (2004), 39-73.
280 B. Molnár

[4] Blokdijk, A. and P. Blokdijk, Planning and Design of INformation

Systems, Academic Press Professional Inc., 1987.
[5] Boman, E., K. Devine, R. Heaphy, B. Hendrickson, V. Le-
ung, L.A. Riesen, C. Vaughan, U. Catalyurek, D. Boydag, W.
Mitchell et al., Zoltan 3.0: Parallel parttioning, load balancing and data-
management services. User guide, Technical report, NM, Sandia National
Laboratories, Albuquerque, 2007.
[6] Bocacich, P., A.C. Holdren and M. Johnston, Hyperedges and mul-
tidimensional centrality, Social Networks, 26 (3) (2004), 189-203.
[7] Breitgand, D. and A. Epstein, Sla-aware placement of multi-virtual
machine elastic services in compute clouds, Integrated Network Manage-
ment (IM). 2011 IFIP/IEEE Int. Symp., IEEE, 2011, 161-168.
[8] Bretto, A., Applications of hypergraph theory: A brief overview, Hyper-
graph Theory, Springer, 2013, 111-116.
[9] McBrien, P. and A. Poulovassilis, A semantic approach to integrating
xml and structured data sources, Advanced Information Systems Engineer-
ing, Springer, 2001, 330-345.
[10] Bruni, R., A. lluch-Lafuente, U. Montanari and E. Tuosto, Style-
based architectural reconfigurations, Bulletin of the European Association
for Theoretical Computer Science, 94 (2008), 161-180.
[11] Cambazoglu, B.B. and C. Aykanat, Hypergraph-partitioning-based
remapping models for image-space-parallel direct volume rendering of un-
structured grids, IEEE trans. on Parallel and Distributed Systems, 18 (1)
(2007), 3-16.
[12] Catalyurek, U.V., E.G. Boman, K.D. Devine, D. Bozdag, R.T.
heaphy and L.A. Riesen, A repartitioning hypergraph model for dy-
namic load balancing, J. of Parallel and Distributed Computing, 69 (8)
(2009), 711-724.
[13] Chiu, C.-M. and M. Bieber, A dynamically mapped open hypermedia
system framework for integrating information systems, Information and
Software Technology, 43 (2) (2001), 75-86.
[14] Corradani, A., U. Montanari and F. Rossi, An abstract machine for
concurrent modular systems: Charm, Theoretical Computer Science, 122
(1) (1994), 165-200.
[15] Erl, T., Service-oriented architecture: Concepts, technology and design,
Pearson Education India, 2005.
[16] Ferrari, C.L., D. Hirsch, I. Lanese, U. Montanari and E. Tu-
osto, Synchronised hyperedge replacement as a model for service oriented
computing, Formal Methods for Components and Objects, Springer, 2006,
22-43.
Applications of hypergraphs in informatics 281

[17] Gamma, E., R. Helm, R. Johnson and J. Vlissides, Design Pat-

terns: Elements of Reusable Object-Oriented Software, Pearson Education,
1994.
[18] Open Group, Togaf: The open group architecture framework, version 9,
http://www.opengroup.org/togaf/, 2010.
[19] Herbsleb, J.D. and R.E. Grinter, Architectures, coordination and
distance: Conway’s law and beyond, IEEE Software, 16 (5) (1999), 63-70.
[20] Larman, C., Applying UML and Patterns: An Introduction to Object-
Oriented Analysis and Design and Iterative Development, Pearson Educa-
tion India, 2012.
[21] MacKenzie, C.M., K. Laskey, F. McCabe, P.F. Brown, R.M.
and B.A. Hamilton, Reference model for service oriented architecture
1.0, 2006.
[22] Marini, J., Document Object Model, McGraw-Hill Inc., 2002.
[23] Molnar, B. and Z. Vinczeller, Comparative study of architecture for
twitter analysis and a proposal for an improved approach, IEEE 4th Int.
Conf. on Cognitive Infocommunications (CogInfoCom), IEEE, 2013, 11-
16.
[24] Molnár, B., A. Benczúr and Á. Tarcsi, Formal approach to a web
information system based on story algebra, Singidunum J. of Applied Sci-
ences, 9 (2) (2012), 63-73.
[25] Molnár, B. and Á. Tarcsi, Design and architectural issues of contempo-
rary web-based information systems, Mediterranean J. of Computers and
Networks, 9 (2013), 20-28.
[26] Chul-Ki Nam, Gil-Sang Jang and Jae-Hak J. Bae, An xml-based
active document for intelligent web applications, Expert Systems with Ap-
plications, 25 (2) (2003), 165-176.
[27] Schewe, K.-D. and B. Thalheim, Reasoning about web information
systems using story algebras, Advaces in Databases and Information Sys-
tems, Springer, 2004, 54-66.
[28] Suh, N.P., Axiomatic Design: Advantages and Applications, Oxford Uni-
versity Press, New York, 2001.
[29] Szucs, I., G. Gombos and A. Kiss, Five ws, one h and many tweets, 4th
IEEE Int. Conf. on Cognitive Infocommunications (CogInfoCom), IEEE,
2013, 441-446.
[30] Dimitrakos, T., J. Martrat and S. Wesner, Service oriented in-
frastructures and cloud service platforms for the enterprise: A selection
of common capabilities in real life business trials by the BEinGRID con-
sotium, Springer, 2009.
[31] Wasserman, S., Social Network Analysis: Methods and Applications, 8,
Cambridge University Press, 1994.
282 B. Molnár

[32] Weinman, J., Cloudonomics: The Business Value of Cloud Computing,

John Wiley & Sons, 2012.
[33] Zachman, J.A., A framework for information systems architecture, IBM
Systems J., 26 (3) (1987), 276-292.

Bálint Molnár
Information Systems Department
Eötvös Loránd University
Pázmny Péter sétány 1/C
H-1117 Budapest, Hungary
[email protected]