Jānis Grabis, Mārīte Kirikova and Jānis Vanags
Faculty of Computer Science and Information Technology, Riga Technical University, Kalku 1, Riga, Latvia
Keywords: Fractal systems, Information architecture, Information systems design.
Abstract: The fractal approach has emerged as a promising method for development of loosely coupled, distributed
enterprise information systems. This paper investigates application of information architecture in
development of fractal information systems. Principles of designing the information architecture of fractal
information systems as well as rules for analyzing the information architecture are developed. These rules
are used to obtain problem-domain representations specifically suited for needs of individual fractal entities.
The usage of the information architecture in implementation of the fractal information system for the
university’s study programme development problem is demonstrated.
Modern enterprises operate in close collaboration
with many other enterprises. Support for networking
and collaboration has become a vital requirement for
information systems (IS). That includes addressing
heterogeneity of systems, dealing with conflicting
objectives, accounting for dynamic changes in the
network, information sharing, knowledge exchange,
and providing different, specialized views of the
system. Despite elaboration of various technologies
(e.g. workflows, groupware, content and knowledge
management, virtual enterprises) addressing some
aspects of these issues, they remain of major
importance in enterprise computing and software
development. A fractal approach is also emerging as
a promising technology for designing multiple-
interrelated systems (Warneke, 1993; Hoverstadt,
2008). In the area of enterprise ISs, the fractal
approach appears suitable for dealing with ISs
development problems characterized by a relatively
loose coupling among entities involved in solving of
focused knowledge intensive problems without
highly elaborated and structured workflows.
Examples of such problems are operations of project
consortiums, academic institutions and distributed
product-design groups.
Fractal systems consist of self-similar, self-
optimizing, goal-oriented fractal (independently
acing organizational entities) arranged in a loosely
coupled hierarchical network. They are continuously
evolving and are characterized by rich information
exchange flows inside fractal entities, between
different levels of the fractal system and with
external environment (Ryu and Jung, 2003). Goal-
orientation allows balancing individual and system-
wide interests of all entities involved. Self-similarity
allows simplifying and structuring design of what
might appear as a chaotic system. Self-organization
allows finding ways for achieving goals without having
predefined processes. Information flows supported by
the fractal systems facilitate knowledge exchange.
Ryu and Jung (2003) and Kirikova (2008) discuss
general aspects of development of Fractal
Information Systems (FIS). This paper focuses on
information structuring and management issues in
FISs by means of elaboration of Information
Architecture of Fractal Information Systems
(IAFIS). Information Architecture (IA) describes the
structure of a system, i.e., the way information is
grouped, navigation methods and terminology used
within a system (Barker, 2005). That is particularly
important in fractal systems because a common,
easily accessible information basis is necessary for
fractal entities to achieve their and system-wide
objectives. Additionally, IA defines information
flows among fractal entities. From the ISs
development perspective, IA is used to develop self-
similar representations of the problem domain for
entities involved in the problem solving and to trace
information interdependencies.
Grabis J., Kirikova M. and Vanags J. (2009).
In Proceedings of the 11th International Conference on Enterprise Information Systems - Information Systems Analysis and Specification, pages
DOI: 10.5220/0001988101500155
Thus, the objective of this paper is to elaborate
methods for designing and analyzing IAFIS and to
demonstrate application of IAFIS in development of
FISs. It is assumed that each entity belonging to a
fractal system has its own problem representation,
which is based on a common goals and ontology of
the fractal systems. This representation suits needs
of a particular entity. IAFIS defines all Information
Elements (IE) characterizing the problem domain
and relationships among these elements. Analysis of
IAFIS yields IEs relevant to the problem
representation of individual entities. The
contribution of this paper is elaboration of rules for
analyzing IAFIS, as well as outlining of principles
for designing FISs on the basis of IAFIS. These
principles also can be applied in design of enterprise
content management systems and portals, which are
enterprise systems having limited support for
formalized development. Design, analysis and
application of IAFIS throughout the paper is presented
by using a problem of ISs development for Study
Programme (SP) development in Latvian universities.
For purposes of this paper, the FIS is defined as a
problem-oriented IS shared by a network of
interrelated Organizational Entities (OE), where
each entity has its own representation of the problem
and information needs. Key elements of the FIS are
shown in Figure 1.
Figure 1: A fractal system and its IS. R
denotes problem
domain representation.
A FIS is created upon demand by a group of
entities involved in solving a common problem. The
common problem is defined by a set of goals and
core ontology. The core ontology defines initially
agreed concepts characterizing the problem domain
(Kirikova, 2009). However, it is likely that each
entity also has its own internal ontology, which to
some extent deviates from the core ontology because
usually it is not restricted just to the particular
problem domain. Entities might be arranged
hierarchically. However, they are relatively
independent, and entities at higher hierarchical
levels provide only problem-solving goals and
general framework while the choice of sub-goals and
particular problem-solving mechanisms is not
strictly regulated. An IS supports problem-solving. It
provides multiple views or representations of the
problem domain. These representations are suited
according to specific needs of each entity. At the
same time, they use the common core ontology and
common design principles. IAFIS is used to provide
a systematic framework for developing and
maintenance of these different representations.
In the FIS, an organization entity uses information to
complete its activities and achieve its objectives. It
either is an owner of this information or consumes
information provided by other OEs. IAFIS defines
Information Elements (IE) and relationships between
these elements. It allows to identify information
needs of each OE and to reason about change and
knowledge propagation inside the fractal system.
UML is used to describe IAFIS.
There are five types of nodes used to define
IAFIS (Figure 2). The central element is
InformationElement. It is used to describe any kind
of information unit (e.g., document, record, file)
relevant to a particular problem domain.
InformationElement may contain multiple
Parameter elements. The Parameter element
identifies data items of the IE what either
characterize this IE or have major importance in the
problem domain. It can be represented either as a
class or as an attribute. Parameters often are
numerical values, which can be used to quantify and
analyze the problem domain. IEs together form a
problem representation suitable for a particular OE.
Multiple OEs can share one representation. One
organizational can have multiple problem
representations. Definitions and meaning of IEs and
parameters are provided in either the core ontology
or ontologies owned by individual OEs.
Six types of relationships among IEs are defined.
Association is used to describe general connections
between IEs. A dependency relationship is used to
Param e ter
Pa rameterization
Figure 2: General representation of IAFIS.
Univ ersity
University: : StudyProgrammeRegulations
- MandatoryCoursesCredits: Parameter
Univ ersity::C ourseRegister
Univ ersity::
Insti tute
ProjectWork Regulations
Institute::Cou rceDescription
Figure 3: IA of SP development information system.
describe some kind of dependence of one IE upon
other IE. It is particularly used to show that content
of one IE is developed according to some
requirements provided by another IE. Aggregation
and composition relationships are used to describe
that an IE consists of other IEs. A parameterization
relationship indicates that a
Parameter element
belongs to the specified IE. A summation
relationship is used to describe data transformation
and aggregation relationships between different
problem domain representation.. IAFIS is developed
using the defined elements. A designer identifies
fractal OEs and IEs, assigns the IEs to the OEs,
parameterizes the IEs and establishes relationships
among the IEs. Thereafter, IAFIS is analyzed and
used in development of problem domain
representations for individual OEs and for
maintenance of the FIS by tracking change and
knowledge propagation.
The example of study program development IS is
used to illustrate IAFIS. SPs comply with general
guidelines set by the law. The SPs are accredited by
the Latvian Ministry of Education and Sciences. On
the basis of these legal requirements, each university
develops their internal regulations on development of
SPs. These regulations are more detailed and include
references to other administrative regulations and
documents (e.g., course register). SPs are developed
by university’s faculties.
A fragment of IA for the SP development IS is
shown in Figure 3. It consists of two OEs (depicted
using parallelogram) – university and faculty (other
hierarchical levels such are omitted). The SP
development at the university is governed by “SP
ICEIS 2009 - International Conference on Enterprise Information Systems
Development Regulations” represented by the
University::StudyProgrammeRegulations IE. There
are multiple data items characterizing this IE, and
these are presented using elements of type Parameter
(only one parameter is shown for the sake of
simplicity). For instance, the regulations mandate
the total number of credit hours required in a SP.
This is an important characteristic and therefore is
shown as a parameter. SPs are developed by
faculties, and IE
Faculty::Study Programme
belongs to the faculty. The dependency relationship
is used to show that the SP is developed according to
requirements provided by the SP development
regulations. The SP also has a parameter
characterizing the total number of credit hours. The
summation relationship is used to describe
information flows from one hierarchical level to
another. In this case, it is shown that the University
sets the bounds on the total number of credit hours
and that the University averages total credit hours
data received from faculties (this information can be
used to analyzed structure of SP and to judge about
necessary adjustments in regulations). The SP IE is
also shown to depend upon
and indirectly upon
University::CourseRegister. There are also IEs,
which are unique to the faculty and do not directly
depend upon IEs belonging to other OEs.
IAFIS is analyzed in order to develop
representations of the problem domain for each OE.
The analysis yields several types of collections of
IEs gathered from different OEs. These collections
are aimed to contain all IEs and their parameters
necessary for an OE to address the particular
problem. The entity specific problem representations
consist of properly arranged collections. IAFIS is
also analyzed to understand information
interdependencies, to identify isolated IEs as well as
to investigate other features of fractal systems.
The most important task of the analysis is
identification of required IEs for each problem
representation. These IEs are grouped in collections.
The first collection C
includes all IEs owned by a
particular OE. The second collection C
includes all
IEs, which are suppliers in a dependency
relationship with IEs belonging to the first group.
The third collection C
consists of all IEs, which are
indirect suppliers of IEs. The fourth collection C
includes IEs, which have any other types of direct
relationships with IEs belonging to C
. These rules
are formally specified using OCL (Object
Constraints Language). Collections of IEs
subsequently can be used during the implementation of
the FIS for grouping and to establish hierarchy of IEs.
IEs in IAFIS are parameterized to highlight the
most important characteristics of the problem
domain. Relationships between parameters are
shown using the summation relationship. The
summation relationship is bidirectional. From a
supplier to a client, it describes what kind of
restrictions the supplier imposes on the client. From
a client to a supplier, it defines the supplier-side
processing of data provided by the client.
Parameterization and summation relationships
are also analyzed. That includes finding all para-
meters characterizing a particular IE. For each OE,
three groups of parameters are identified. The first
group P
includes parameters characterizing each IE
from collection C
. In the FIS, these are displayed
along the particular IE. The second group P
includes parameters directly or indirectly provided
by clients in the summation relationships. These can
be used as key performance indicators. The third
group P
includes parameters directly or indirectly
provided by suppliers in the summation
relationships. These can be used as the most
important problem-solving guidelines.
The analysis of IAFIS is also used to update the
ontology of the fractal system. IEs and parameters
used by OEs are matched against concepts defined
in the core ontology. If these elements are not found
in the core ontology, they are either identified as
candidates for inclusion in the core ontology or
inspected for correspondence to concepts already
included in the core ontology. An OE specific
problem representation contains only those elements
deemed explicitly necessary to problem-solving
though tracing capabilities also can be provided.
The analysis also is used to identify isolated IEs
and isolated clusters of IEs. These collections are
candidates for knowledge propagation and
modification of the core ontology in the case of
semantical inconsistencies.
During maintenance of the FIS, IAFIS is updated
both automatically and manually in response to
changes in the fractal system and problem-domain.
The updating is classified as change propagation and
knowledge propagation. The FIS can be modified
according to the changes made in IAFIS.
Figure 4: Sample implementation of SPDIS.
Change propagation deals with updating of IA in
the case of changes in its elements or relationships.
Three kinds of change propagation situations are
considered: 1) updating in the case of added IE; 2)
updating in the case of added parameter; and 3)
updating in the case of added relationships.
In the case of added IE, relationships between
the element and other elements owned by the
particular entity are manually established. The
ontology of the particular entity is also updated. If
the element is added at upper levels of the fractal
system, semantically related IEs belonging to lower
level entities are searched in the core ontology, and
the lower level entities are notified to consider
updating of their representations. Adding an element
at lower level entities often is performed in response
to changes in upper levels and dependency
relationships can be established. If the element
initially is added for internal used, it is defined in the
ontology of the particular entity and its further
evolution depends upon rules of knowledge
In the case of added parameter, summation
relationships are added to the IA. Initially, all
parameters of directly or indirectly related IEs in all
other representations are checked to identify
semantically related parameters. The core ontology
is used in the identification process. Summation
relationships are established with semantically
related parameters. If semantically related
parameters are not found, a new parameter is added
to related IEs, a new IE and its parameters are added
or no action is taken. If new elements are added then
summation relationships are also established. In the
case of added relationships, there are no direct
changes in IAFIS but the analysis rules are
reevaluated and changes are resembled in the FIS.
Knowledge exchange is vital for fractal systems.
Knowledge can be propagated from individual
entities to the whole fractal system. Three types of
knowledge propagation mechanisms are identified:
1) promotion of IEs with high level of cohesion
); 2) best practice propagation (K
); and 3)
promotion of frequently used elements (K
). From
the IA perspective, an IE is of type K
if it is
involved in many direct or indirect relationships
what indicates that this element is important to the
problem domain. In the case of human directed
knowledge propagation, representations lacking
elements of type K
are analyzed to check relevance
of these elements. Automatically, elements from K
can be provided as recommendations (Montaner et
al., 2003) for inclusion in the representation.
The fractal system adopts best practice processes,
which have been successfully utilized by some of
fractal entities similarly as described by Steckuka et
al. (2008). IEs used in these processes are included
in problem-domain representations for those entities
adopting the processes. IEs of type K
determined by monitoring usage of the FIS. The
fractal system automatically recommends adding to
the representation IEs frequently used by other OEs
or IEs frequently requested using the trace function.
Similar knowledge mechanisms can also be
applied for propagating knowledge about
ICEIS 2009 - International Conference on Enterprise Information Systems
Application of IAFIS is demonstrated by
implementing a prototype of the information system
for the SP development problem (SPDIS). The IS is
implemented on the basis of commercial colla-
boration and content management system according
to IAFIS shown in Figure 3. Figure 4 shows the user
interface of SPDIS of the university. Markers are
used to indicate different parts of SPDIS. The first
part refers to problem domain representations for
different OEs (e.g., university and faculty). Part 2
contains links to IEs needed by the particular OE.
These elements are collected and structured accor-
ding to the rules established in Section 5.1. Part 3
contains the selected IE, in this case the regulations
on SP development. Part 4 lists parameters of the
selected IE. The list contains their title and value,
and summation value, which is computed from data
provided by clients in the summation relationship.
Part 5 lists all IEs from the collection C
. The
recommendations part (part 6) demonstrates
automatic knowledge propagation. The IEs in this
part are included according to the rules K
and K
specified in Section 6.2. Part 7 shows parameters,
which are used by the university in elaboration of
regulations on SP development. The parameters are
those included in group of parameters P
The paper has proposed IA and its analysis rules as a
tool for developing fractal systems. The problem
representations specifically suited for particular OEs
and built on the basis of IAFIS are self-similar what
ensures consistency in the relatively loose coupled
system and reduces systems development and
maintenance cost. At the same time, they are
adjusted to needs of particular OE, which have the
sufficient information basis to complete their tasks
with respect to common and individual goals. The
set of rules for change and knowledge propagation
enables updating of the FISs and facilitates
knowledge sharing among fractal entities. To our
knowledge, the proposed IA and its analysis rules
provide the first systematic framework for
information management in fractal systems.
Efficient utilization of IAFIS requires
parameterization of IEs. Concept modeling and
document mining techniques can be used for this
purpose. The fractal system can be designed in either
top-down or bottom-up manner. In the case of top-
down approach, a lead entity develops its problem
representation and this representation can be used as
a template for developing self-similar represent-
tations. In the case of bottom-up approach, fractal
entities have their own problem-representations,
which are continuously aligned during evolution of
the fractal system. Another question for future
research is integration of fractal systems with other
ISs because IAFIS and implementation of FIS
depends upon already existing models and systems.
Adomavicius, G. Tuzhilin, A., 2005. Toward the next
generation of recommender systems: a survey of the
state-of-the-art and possible extensions. IEEE
Transactions on Knowledge and Data Engineering 17
(6), 734–749.
Barker, I., 2005. What is information architecture?
available at
Hoverstadt, P., 2008. The Fractal Organization: Creating
Sustainable Organizations with the Viable System
Model, Wiley, New York.
Kirikova M., 2008. Towards multifractal approach in IS
development. In Barry, C. et al. (eds). The Inter-
Networked World: ISD Theory, Practice, and
Education, ISD2007, Galway, Ireland, August 29-31,
2007, Springer-Verlag: New York.
Kirikova, M., 2009. Towards flexible information
architecture for fractal information systems. In The
proceedings of the Int. Conf. on Information, Process,
and Knowledge Management, eKNOW.
Montaner, M., Lopez, B., De la Rosa, J.L., 2005. A taxonomy
of recommender agents on the internet. Artificial
Intelligence Review 19 (4), 285–330.
Ryu, K., Jung, M., 2003. Fractal approach to managing
intelligent enterprises. Creating Knowledge Based
Organisations, Gupta, J.N.D., Sharma, S.K. (Eds.), Idea
Group, 312-348.
Stecjuka J., Makna J., Kirikova, M., 2008. Best practices
oriented business process operation and design. In
Proceedings of the BPMDS'08 Workshop, 9th
Workshop on BPM, Montpellier, France, June 16-17,
2008, 171-184.
Warneke, H.J., 1993. The Fractal Company, Springer
Verlag, New York.