Knowledge-Based Systems 14 (2001) 259±262
www.elsevier.com/locate/knosys
Information visualization for intelligent decision support systems
Tong Li a,*, Shan Feng a, Ling Xia Li b,c
a
Department of Automatic Control Engineering, Huazhong University of Science and Technology, Wuhan 430074, China
b
Department of Information Systems and Decision Sciences, Old Dominion University, Norfolk, VA 23529, USA
c
National Key Laboratory for Manufacturing Systems Engineering, Xian 710049, China
Abstract
To work ef®ciently with decision support systems (DSS), most users bene®t from representation conversion, i.e. translating the speci®c
outcome from the DSS, normally portrayed in a numerical format, into the universal language of the visual. In general, interpretation of data
is much more intuitive if the results from the DSS are translated into charts, maps, and other graphical displays because visualization exploits
our natural ability to recognize and understand visual patterns. In this paper we discuss the concept of visualization user interface (VUI) for
DSS. A proprietary software system known as AniGraftool is introduced as an example of an information visualization application for DSS.
In addition, a visualized information retrieval engine based on fuzzy control is proposed. q 2001 Elsevier Science B.V. All rights reserved.
Keywords: Information visualization; User interfaces; Fuzzy sets; Software tools; Decision support systems
1. Introduction
In modern decision support systems (DSS), an increasingly larger percentage of the total design effort is devoted
to the user interface, or that portion of the software system
concerned with providing the means for a human user to
interact with a system's application software. The structure
of the user interface therefore has a major impact on the
quality of the whole system [1±5].
Current DSS user interfaces are dominated by the presentation of information via character and numerical formats
and is often considered an area of pursuit for those who have
a propensity for numbers and enjoy mathematical equations.
Unfortunately, many of the individuals who stand to bene®t
most from DSS, such as CEOs, executives, politicians, administrators, etc. lack this mathematically-oriented mindset.
To work ef®ciently with DSS, most users bene®t from a
ªrepresentation conversionº, i.e. translating the speci®c
DSS alphanumeric results into the universal language of
the visual. In general, interpretation of data is much more
intuitive if the results from the DSS are translated into
charts, maps, and other graphical displays because visualization exploits our natural ability to quickly recognize and
understand visual patterns. For instance, family planning
DSS users better understand the problem of population
age structural impacts through the display of an intuitive
visually oriented population age pyramid. Macro-economic
DSS users, for example, have a better grasp of the momen-
* Corresponding author.
tum of national industrial structure after seeing a bar chart
moving up or down dynamically.
Steen summed up this problem nicely by de®ning the
expression ªI seeº in relation to mathematics ([6]): ª'I
see' has always had two distinct meanings: to perceive
with the eye and to understand with mind. For centuries
the mind has dominated the eye in the hierarchy of mathematical practice; today the balance is being restored as
mathematicians ®nd new ways to see patterns, both with
the eye and with the mind.º
This can be also applied to DSS. Information visualization
exploits the natural human ability to recognize and understand
visual patterns. For many people, it is the easiest and most
intuitive way to interpret data in the DSS application domains.
2. Visualized user interface (VUI) for DSS
In the 1970s, the user interface for DSS consisted primarily of a one line at a time dialogue with the computer, either
through a command language, or through a question-answer
style format. Beginning in the mid 1980s, DSS user interfaces had evolved to a mouse-driven, multi-windows interfaces such as that found in Apple's Macintosh computers,
Microsoft's Windows system and X Window based
systems. Visualized User Interface for DSS (VUI-DSS) is
the next step in the evolution of DSS user interfaces.
The goal of information visualization was mainly to
provide suitable methods and instruments to explore and
depict data and information through graphical representation. Information visualization takes advantage of the fact
0950-7051/01/$ - see front matter q 2001 Elsevier Science B.V. All rights reserved.
PII: S 0950-705 1(01)00104-6
260
T. Li et al. / Knowledge-Based Systems 14 (2001) 259±262
Fig. 1. Visualization pipeline.
that visual representations can serve as powerful ªvehicles
of thinkingº that help us extract useful information from
complex and/or voluminous data sets [7]. It also provides
processes for manipulating the data set and seeing what may
have previously been invisible, thereby enriching existing
investigation methods. However, the concept of information
visualization is no longer limited to the graphical display of
data but now encompasses a much broader spectrum including the design of graphical interfaces used to input and
access that data, in addition to the creation of standard
and novel data presentation formats. With the overwhelming amount of information that is generated and received
through OLTP, well-designed vehicles for facilitating data
capture coupled with creative and powerful means for
clearly, accurately and concisely conveying meaningful
information are essential to effective DSS implementation.
The DSS designers should offer users effective solutions for
accomplishing these tasks. From our point of view, information visualization is not a goal in itself, but an integral part
of the overall process of presenting scienti®c data.
Information visualization functions to support DSS therefore have to be embedded into the system's software that
deals with all the aspects of the scienti®c problem.
As Fig. 1 shows, the information visualization model is
referred to as the ªvisualization pipelineº and consists of
®ve stages: simulating, preparing, mapping, rendering and
interpreting [8].
Data ¯ows from left to right through the pipeline. The
ªSimulating Stageº is the starting step. The ªPreparing
Stageº involves data preparation through normalization or
other mathematical steps. The ªMapping Stageº represents
the natural juncture of scienti®c and graphical data and as
the key to the visualization pipeline, is the most dif®cult
step. The ªRendering Stageº and ªInterpreting Stageº are
the last two steps. VUI for DSS addresses the mapping stage
of the visualization pipeline and is essentially a data- and
knowledge-based graphical software shell that automatically translates DSS outcome data into charts, maps, and
animations.
3. AniGraftool: an example of VUI for DSS
3.1. Tools and applications: anigraftool and DisMEIDSS
AniGraftool is a software development tool designed for
VUI for DSS that aids in the development of visual user
interfaces for DSS applications software. Display for Macro
Economic Intelligent DSS (DisMEIDSS) is a graphical
display with animation features for MEIDSS, an intelligent
decision support system for macroeconomic decisionmaking [9,10]. AniGraftool has been employed in
DisMEIDSS. With a click of the mouse, the software automatically translates the resulting data from the DSS into
charts, thematic maps, scatter plots, and other graphics.
As an example, time-series data can be translated into
ªmovingº charts or thematic maps which dynamically
change their shape or color-coding. DisMEIDSS's other
features include its capacity for knowledge-based query
and animation. The information in application domains for
DSS can be inquired through indices.
3.2. Basic elements
The AniGraftool development software consists of three
elements: (1) an interpreter/inference engine (an executable
that runs under MS-Windows software); (2) a set of knowledge base ®les, including script ®le, keyword de®nition ®le
(standard ASCII text ®les); and (3) a set of data base ®les
(space-separated standard ASCII ®les) (see Fig. 2).
3.2.1. Interpreter/inference engine
The interpreter/inference engine is an MS-Windows
compatible program with Multiple Document Interface
(MDI) style written in Borland C11. During run time it
reads knowledge base (KB) and data base (DB) ®les. There
are three types of KB and DB ®les: (1) script ®les (ending
with p .lot extension); (2) data ®les (ending with p .dat);
and (3) keywords de®nition ®le (ending with p .def) for
information query. These KB and DB ®les make up a
speci®c application.
The purpose of the interpreter/inference engine is to
search the users' needs by his/her query and to translate
the operators in the KB and DB ®les into executable functions and routines for displaying a speci®c dialogue box,
loading certain sections of the data, displaying a speci®c
chart, or running an animation sequence. The interpreter/
inference engine gets its commands from the KB and DB
®les through the user's selection or his/her questions. They
precisely tell the interpreter/inference engine how the data
are organized and what it should do with them.
3.2.2. Script ®les
The script ®les make up the core of the software. They are
linked to menu or dialogue box items and precisely de®ne
what the interpreter/inference engine should do in response
to the user clicking on that speci®c menu or dialogue box
item. For instance, the script ®les de®ne which columns/
rows the interpreter/inference engine should read from a
data ®le. One can specify coherent blocks of data or very
complex non-rectangular data structures. One can also use
the data description to select columns/rows in a data ®le that
should be used for animation, etc. Besides this simple (but
quite powerful) data description language, the script ®le has
sections, which de®ne the type and layout of the screen
display. The developer can specify various kinds of graphs
T. Li et al. / Knowledge-Based Systems 14 (2001) 259±262
261
Fig. 2. Main elements of the AniGraftool.
(e.g. bar, line, pie, area, scatter plot, maps, etc.). All charts
and maps can also be animated.
3.2.3. Data ®les
The data ®les are organized in very simple format: they
are space-separated standard ASCII ®les which makes it
easy to prepare the data for the graphical database. In the
®rst section there can be a data header, such as the title of
the speci®c data set, etc. that readily identi®es what it is
about. The next section is the data section, which are
organized into rows and columns. A more advanced data
structure could be envisioned than this simple, spreadsheettype arrangement, but we think that this scheme has four big
advantages: ®rst, it is easily interactive with the application
portion (the simulating stage) of the software system. The
speci®c outcome from the DSS could be easily processed to
get the scheme by a simple data preparation program.
Second, it is intuitive. Most computer literate individuals
can immediately work with data ®les that stick to the
column/row concept. Third, it is simple. It does not require
understanding of more advanced record-oriented data structures. Fourth, One can observe the data ®les directly. It is not
necessary to run special software just to look into the data
®le (as is the case with many data bases distributed on tape).
Since it is a plain ASCII ®le, any editor can be used to check
the data.
3.2.4. Keyword de®nition ®les
The keyword de®nition ®le is for knowledge-based information query. Some de®nite concepts of the special application domain for a DSS are de®ned as the keywords for the
indexes of the information query.
3.3. Knowledge-based information query
The development of a software tool to support the user
interface design of a DSS is not suf®cient by itself. In addition, we need to enhance the human operator's ability to use
these tools. For example, although information can be
displayed graphically, users may not be able to understand
all of what's being displayed or may be overwhelmed if too
many graphical displays are presented on the same screen.
To minimize these shortcomings, a fuzzy control engine is
embedded in the AniGraftooldevelopment tool that supports
a fuzzy query based on speci®c keywords in the application
domains.
As we know, the input of a computer application system
consists of a given set of symbols. The output is a set of
symbols that is readily understandable by the user. A good
system, therefore, should support users in ®nding meaningful information in a simple way, such as through inputting
certain keywords. A fuzzy retrieval engine based on the
262
T. Li et al. / Knowledge-Based Systems 14 (2001) 259±262
that eventually better decisions can be reached. In this
paper, we have discussed the role that information visualization plays in the DSS and introduced the software development tool AniGraftool as an example of the VUI for DSS.
Information visualization is a powerful tool with tremendous potential for supporting complex decision support
problems and their problem-solving processes. However,
information visualization is still in its infancy and requires
advanced study on techniques and applications.
Fig. 3. Fuzzy retrieval engine and query in Anigraftool.
concept of fuzzy control that is supported by the system has
been included (see Fig. 3).
Many decision support problems have data that lack
obvious structures to provide information visualization
with a base. We formulate Y WF X: Y is the set of
symbols representing the user's need (also called the satis®ed solution set). X is the input, such as some keywords
speci®ed in our system. F is a transformation function that
transforms X into F(X). ^ is the ªsatisfactionº operator
which transforms F(X) into Y following the users requirement.
The term ªsatisfactionº is a fuzzy concept and can be
represented by a fuzzy set. Concretely, one to four keywords
that are related to the application domains are allowed to be
entered into the system, which de®ne the scope of information retrieval. The visualized information that is related to
the keywords are provided by the fuzzy retrieval engine (see
Fig. 3). The fuzzy retrieval engine, as a function module
embedded in the system, retrieves the information through
interaction with user. The `item of information' in Fig. 3 is a
classi®ed set of the information about the DSS application
domains. It is the title of the content of the information
supported by one to four keywords about the DSS application domains.
4. Conclusions
Information visualization is an approach that can assist
DSS users in gaining insight into the quantitative data so
Acknowledgements
The project is partially supported by the grant
No.69674041 from the National Natural Science Foundation of China (NSFC). The authors are indebted to Pan
Wang, Jingping Huang, Chao Tang and Thomas Howard,
who have contributed to this work. The authors are also
grateful to the referees for their valuable comments on
this paper.
References
[1] G.O. Domik, Guidelines for a curriculum in scienti®c visualization,
Computers and Graphics 17 (2) (1993) 185±191.
[2] C. Douligeris, et al., Development of OSIMS: an oil spill information
management system, Spill Science and Technology Bulletin 2 (4)
(1995) 255±263.
[3] L.A. Jesse, J.K. Kalita, Situation assessment and prediction in intelligence domains, Knowledge-Based Systems 10 (1997) 87±102.
[4] P. Zhang, An image construction method for visualizing managerial
data, Decision Support Systems 23 (1998) 371±387.
[5] T. Li, S. Feng, P. Wang, L. Xu, Visualization and decision support
systems, in: Proceedings of the 14th World Congress of International
Federation of Automatic Control, IFAC, Beijing, 1999.
[6] L.A. Steen, On the Shoulders of Giants: New Approaches to Numeracy, National Academy Press, Washington, DC, 1990.
[7] R.H. McKim, Experiences in Visual Thinking, PWS Engineering,
Boston, MA, 1980.
[8] D.S. Dyer, A data¯ow toolkit for visualization, IEEE Computer
Graphics and Applications 10 (1990) 60±69.
[9] S. Feng, L. Xu, Decision support for fuzzy comprehensive evaluation
of urban development, Fuzzy Sets and Systems 105 (1) (1999) 1±12.
[10] S. Feng, L. Xu, Hybrid arti®cial intelligence approach to urban planning, Expert Systems 16 (4) (1999) 248±261.
