Download Organisational Decision Support System Vs. Data Mining Methods

Organisational Decision Support System Vs. Data
Mining Methods pertinence to estimate and manage risk
of innovative projects
Kamal Touati, Fahima Nader, Anne-Marie Alquier1
1
Département de Sciences Sociales pour l'Ingénieur, Université Toulouse 1,
Place Anatole France, 31042 Toulouse Cedex, France
[email protected], [email protected], [email protected]
Abstract. This paper presents research foundations and results of the PRIMA
project (Project RIsk MAnagement, IST-1999-10193, 2000-2001).
The project proposes a generic approach to design and develop an
Organizational Decision Support System (ODSS). The type of ODSS presented
(risk estimation and management of innovative projects made during the
bidding phase) is generic among process-oriented organizations.
The proposed approach is pulled by requirements and its steps follow up
referenced Management Information Systems (MIS) design and development,
starting from requirements engineering, then defining the knowledge content
necessary to decision activities and finally using easy-to-develop tools. KDD
techniques belong to these tools.
The comparison of this approach with KDD methods using the company
information systems is made, and the conclusion is that no overlap appears.
Nevertheless, later use will have to be considered, which is a starting research.
1
Introduction
This paper presents research foundations and results of the PRIMA project (Project
RIsk MAnagement, IST-1999-10193, 2000-2001). The project proposes a generic
approach to design and develop an Organizational Decision Support System (ODSS).
The type of ODSS presented (risk estimation and management of innovative projects
made during the bidding phase) is generic among process-oriented organizations [4],
[9].
The decision situation is explained first. The specific type of Decision Support
System aimed at improving this situation is presented afterwards.
ODSS definition and their specific problems of design and development are
demonstrated.
The generic approach to design and develop an Organizational Decision Support
System (ODSS) is shown. This approach is pulled by requirements and its steps
follow up referenced Management Information Systems (MIS) design and
development [10], [11], [16], starting from requirements engineering, then defining
the knowledge content necessary to decision activities and finally using easy-todevelop tools. KDD techniques belong to these tools.
The comparison of this approach with KDD methods using the company
information systems is made, and the conclusion is that no overlap appears.
Nevertheless, later use and connection will have to be considered, essentially when
return on experience from projects is organised. But this connection is not just to
reuse the ODSS indicators as the information to be expected from the KDD process.
A research has therefore been launched to answer this problem.
2 Decision situation : estimate and manage risk of innovative
projects
This paper introduces research foundations and presents the PRIMA project
results. Nine European Partners (France, Spain, Greece & Italy) participate to the
project, which allows to consider main industrial preoccupations and standards (from
Satellites to chemical plants).
The PRIMA project objectives are to develop a Decision Support System (DSS)
applied to risk management of innovative projects. In an era characterised by short
product life-cycles, dynamic markets and complex processes, develop new products
(or services) is becoming the primary source of sustainable competitive advantage.
Competitive advantage is taken into account by focusing on early phases of
projects or product life-cycle. The PRIMA DSS provides an up-front analysis method
for project managers made during the early phase of a project (bidding phase). This
phase is considered as very important in term of return on investment for users. In the
PRIMA project, return on investment is evidenced, at the very beginning of a project,
with analysis and improvement of bid winning chances and of company strategy on
profit.
Sustainable competition is linked to risk management and introduction of new
attitudes about risk in the company, as:
• Risk management can be considered as the most powerful driver to manage
innovative projects. Therefore, the prima project main innovative points are to
consider risk at the highest level and as soon as the contract negotiation. Risk
integrates all other preoccupation's like performance, schedule, cost, affordability,
consumer's needs and priorities, environment,... Risk identification and estimation
is made at business level, contrasting with current approaches where risk is
considered much later in the project, as in liability, safety,...
• To focus on management decisions at the very beginning of a project can be
considered as the most powerful lever to introduce new attitudes about risk in the
company, which is a most important but difficult change. Change starts up at the
upper level of the company stakeholders attitudes and interest [2], as profitability is
enhanced by concentrating on more opportune and more competitive bids.
The DSS is based on a knowledge approach. Risk knowledge is particularly
difficult to organise in a company, as it is most often fuzzy, distributed, unstructured,
tacit, insufficiently organised and catalogued, underestimated, registered in
heterogeneous information systems, secretive, and possessed by experts who are rare
in the company. Knowledge modelling allows to build a corporate memory about
risks, providing. Data extracting methods and knowledge discovery techniques are
attached to this memory to provide the DSS with knowledge support necessary to the
decision making activities of multiple users.
Preparation and negotiation of innovative and future projects can be characterised
as a strategic-type decision situation, involving many uncertainties and an
unpredictable environment. As such, the DSS is the support of a major Project
Management Key Practice [4][9], allowing to organise a very well documented upfront analysis for deciding whether to go into development phase and with which
precautions, covering and control measures. The DSS gain is mainly to improve bid
quality and efficiency, helping:
• to make early decisions, like bid/no bid or make or buy decisions, as recurrent
elements can be used to promptly assemble blue print or sketches,
• bid construction, allowing to examine more alternatives,
• analysis and comparison of the solutions built, and notably identify risk drivers and
isolate innovative knowledge on which the attention must be focused.
• to identify and decide, early in the life-cycle of the product, actions to seize
openings or to avoid and cover major risks (insurance, guarantee,...) susceptible to
endanger the future product,
• the preparation of the project itself once the contract is signed. A risk action plan
(performance indicators and control scoreboards) can be developed, based on the
risks and actions the bid process showed up already.
Quantitative and qualitative impacts of the DSS are to be actually measured:
reduction in the preparation time of a proposal and decrease for proposal costs,
reduction of human resources to deliver a proposal, rapid building of more accurate
bids by recurrent and reverse engineering, improvement of technical excellence by
greater focusing on key issues detected by solutions estimation and comparison,
substantive improved rationality of the risk management process.
3
Specific type of DSS : ODSS
The PRIMA DSS is a specific type of DSS, i.e. an Organisational Decision Support
System (ODSS), and the approach of the PRIMA project is tailored for the specific
problems of designing and developing ODSS.
An ODSS (Organisational Decision Support System) supports and organizes
division of labor in decision-making inside a firm. It focuses on an organizational
process which cuts across organizational functions and hierarchical layers [2]. It
supports interrelated but autonomous local decisions, but its main help is to
coordinate these multiple local decisions with the objective of optimizing
organizational decision. It affects therefore the management level of the company,
introducing a process view and work organization of a firm or even of a virtual
organization including various companies.
It has to satisfy both individual and organizational levels.
At individual level, it has to:
• satisfy multiple types of decision makers, providing individual benefits -for
example in the case of PRIMA: technical managers, commercial managers, project
managers, project management managers,
• improve individual decision models: faster and better identification of problems,
multiplication of alternatives examined, and choice upgrading,
• change individual roles in the organization.
At organizational level, it has to support basic business processes [2], and :
• improve coordination and effectiveness of interdependent decisions.
• support company policy by standardizing guidelines and procedures across the
organization and streamlining organizational business processes
• affect business directly: improving profits, increasing market share and return on
investments, etc.
An ODSS shares characteristics with other management information systems
[10][6], such as DSS, GDSS and EIS, but it has distinctly different objectives and a
broader scope. It has a strong organizational component not present in a DSS or a
GDSS and a coordination component not present in an EIS. Hence, compared to other
management information systems, an ODSS has different functions and components,
and requires different design and development approaches.
Comparison of ODSS with DSS, GDSS and EIS will be considered to find
similarities and isolate specificity's. Comparison between ODSS and EIS will be
specifically stressed as both are supposedly handling mainly techniques and tools of
data mining and knowledge discovery.
ODSS and DSS:
ODSS have to support autonomous decisions and enhance performance of
individual decision-makers. Their design process have therefore common factors with
traditional DSS, and notably the importance given to the cognitive process of the
decision-maker. But ODSS are not just an assemblage of DSS. They are not designed
to support many decisions of one individual decision maker or many independent
decisions of individual decision makers. They support interdependent decisions made
by many individuals with multiple interests. Therefore, users being diverse and
numerous, and coordination among various units being a higher preoccupation,
individual users requirements are not completely satisfied. Individual users roles are
more portrayed conventionally, and user participation from the very beginning and all
along the design is not the utmost rule. Individual users may therefore find ODSS
more impersonal and less relevant than an individually designed DSS.
ODSS and GDSS:
GDSS (Group Decision Support Systems) [3][15], are designed to support decision
making of a group of people (a team) engaged in a decision-related task. GDSS are
supposed to reduce communication barriers, stimulate or hasten exchange of
messages, reduce uncertainty or noise in group's decision process, and drive or
regulate the group's decision process.
GDSS technologies are mainly blackboard-type tools, electronic boardroom, audiovisual conference rooms, group network. From the point of view of knowledge
modelling, the main point is to organise the group information centre as a "group
memory", which provides uniform and consistent knowledge to the group. This
information centre is the basis for people learning from the group.
ODSS enhance also performance of working groups. But if GDSS focus on single
work teams with little differentiation in roles and relationships regulations, ODSS
objectives are to facilitate the interaction of multiple groups, differentiating formally
their roles and relationships, and organising regulation mechanisms. GDSS have to
consider social factors that influence group behavior. ODSS have to consider
organizational factors that influence enterprise performance and behavior.
Organizational-level decision processes involve issues of greater consequence than
group level processes. In ODSS, organizational factors are actually a model of the
global work organization of the company.
Hence, an ODSS cannot be viewed as a simple extension of GDSS, just as group
support systems cannot be viewed as simple extension of individual DSS.
ODSS and EIS:
Executive Information Systems (EIS) are relevant to wide-ranging decisions made
by top executives. They support diverse mix of decisions executives make. As such,
they are not restricted to any particular function inside the company.
Even if they are built and maintained by professional developers, because mainly
lack of time for executives, the corresponding computerised systems may have
relatively simple modelling capabilities. Data mining and EIS-software are mainly
directed at this type of management information systems, as EIS need both :
• easy access to a large number of internal and external information sources relevant
to executives' critical success factors
• and customised presentations which help interpretation by the decision maker.
Common Traits between EIS and ODSS are:
• direct use by top-level executives (ODSS are also directed at other users),
• access to varied sources, both within and outside of the organisation,
• integration of critical success factor or key indicator information,
• and ability to do status reporting, exception reporting, trend analysis, and drilldown investigation.
An already known trend of EIS is to allow lower level managers to get information
consistent with top executives and therefore access in some way to EIS, which is a
propensity to make it an ODSS.
Generally, requirements engineering is made through a Critical Success Factor
method. The method first identifies executive goals through executive interviews.
Afterwards, information that underlie them is formalised: goals are measured through
activities in which satisfactory results will ensure organisational competitiveness;
these activities are aggregated in measure/report progress on goals, with both
objective measures and subjective assessments.
Information sources are external (e.g., customers) and co-ordinated from diverse
internal sources. Information is both about current results (short-run performance), as
well as building for the future.
In conclusion, an ODSS provides critical information to managers like an EIS. The
objectives and scope of ODSS and EIS are however very different. The purpose of an
EIS is primarily to meet the "information needs" of managers, while an ODSS has to:
• provide knowledge sharing
• support decisions of varied users and varied users
• support organizational decision processes and interdependent task execution.
It provides therefore coordination mechanisms to ensure that organizational
decision processes are optimized; for example decisions that can be considered good
at the individual level can be organizationally inappropriate.
4
ODSS architecture and design
4.1
Architecture
The PRIMA DSS architecture is generic and specifically tailored to ODSS. It is
relevant for the conceptual level (requirements engineering) as well as the technical
level design (software engineering).
Conceptually, external and internal risks are separated in two quasi-independent
analysis sessions, with differing knowledge support and processing:
External risk is the risk that the company does not control. They are related to
factors external to the company, arising in the company environment: market shifts,
government action, product interactions with the environment (environmental
protection, regulation context), market competition, use of the product and product
interactions with the customer after product release, external constraints (like
regulation, legal context, currency fluctuations, customer's country regulation
mechanisms and instances). External risk evaluation corresponds to the risk the
company has to face. It is called the market or environment risk.
Internal risk is the risk that is supposed to be under the company control. It is
associated to the technical solutions under analysis during the bid process. It is the
manufacturer's risk (i.e. industrial or technical risk), about its products, processes and
resources : new technology, resources needed for the project or product (partners,
components), processes.
Each analysis session operates a corporate memory specifically oriented on
business, called Risk Management Corporate Memory (RMCM), which supports risk
knowledge processing.
Decision-making activities use and balance external and internal analysis sessions,
with a value analysis type of method.
Intelligence
RMCM: External
information
Design
External risk
Assessment and
ranking
Choice
External value
Data
Data
RMCM: Internal
information
Technical solutions,
cost estimating,
Internal risk
Assessment and ranking
Models Models
Decision-making
support
Internal value
Fig. 1. Generic architecture of the ODSS
The RMCM is operated with data mining tools. But the ODSS modelling method is
quite different from knowledge discovery and data mining methods [5]. The ODSS
modelling process is inverse: modelling starts from use and not data, and useful data
do not pre-exist. Knowledge use (decision support) is therefore the first step, and the
necessary knowledge sources (RMCM) are defined afterwards. Knowledge necessary
to the RMCM is not present in the company, but is co-constructed and emerges from
the RMCM presence.
4.2
Design
The ODDS modelling process will be presented following the three phases of H.A.
Simon’s [12] definition of problem solving: C or “Choice phase”, D or “design
phase” and I or “Intelligence phase”.
ODDS modelling process starts with the Choice phase of the decision activities,
which means that the ODSS modelling begins with the analysis of how the decisionmaker chooses between alternatives. The dashboard(s) uses patterns which are global
assessment indicators of a project [4]: objectives, time, costs, quality, human
resources, performance, processes and risk. The DECIDE project (DECIsion support
for optimal biDding in a competitive business Environment, ESPRIT n° 22298, 9698) identified cost and price indicators, PRIMA focuses in complement on global
internal and external risk exposures.
4.2.1
The EPPMR (Enterprise Product and Process Modelling Repository)
modeling
The EPPMR module assists users during the task of designing technical solutions.
There is tough negotiation between the actors involved in the bidding process and
the customer. The negotiation is centred on the product description and
decomposition and its functionalities. A product is decomposed into entities and each
entity can be decomposed into other entities. The actors are accustomed to this
decomposition principle.
The EPPMR modelling consists of two points : the data base and the model base.
The data base
The actors are accustomed to arguing on products or sub-products, on associated
processes and on the required resources to perform this process. This leads us to
Product*Process*Resource modelling.
A product, a process or a resource could be decomposed into sub-products that are
called Entities, could be decomposed into other processes or resources. This leads us
to the decomposition principle.
Therefore the data representation consists of two aspects :
• the decomposition principle
• and product*process*resource modelling.
Moreover, these two principles allow us to model the “technical aspect” of the
corporate memory, the product or/and the proposal.
The data base model consists of three main entities with a reflexive relationship for
each. This reflexive relationship allows us to treat the decomposition problem. There
is a relationship between each entities.
The three entities are :
• the entity defined with an identification, a name and all the necessary information
to model the product. It performs the entities and associated sub-entities. This
information comes from organisational, technical and costing points of view.
• the process defined in the same way as the entity. It is the process required to build
a product or to assemble different sub-entities.
• the resources needed to perform the process. It is defined by an identification, a
name and other information.
The Data Base Conceptual Model is the following :
Costs
is
decomposed 0,n
0,n
Functionalities
Entity
is
decomposed
0,n
0,n
Process
is
posed
decom
0,n
Resources
N°Proc
- N°Entity
is 0,n!
0,n is 0,n- N°Res
0,n
N
ame
- Name
- Name
linked - ...
linked - ...
- ...
0,n
Unitary
Costs
Fig. 2. The EPPMR Data Model
This model aggregates the two aspects of our problem : the decomposition
principle and the product*process*resource modelling. This model is generic and can
be used in fields other than product manufacturing.
As pointed out previously, this conceptual model is used for :
• modelling the corporate memory and more particularly the design history
• designing the technical solution which remains the predominant use of this model.
The actors will use the system and consequently this modelling the most often for
designing technical solution.
In the first stage of customer negotiation, the users will argue on the product and
entities. Then they will examine the processes matching processes to entities, and
finally the resources assign them processes for technical and cost evaluation.
Therefore the modelling input is the product and its possible decomposition.
The model base
The DSS model base is a model which makes it possible to solve the knowledge
capitalisation and reuse problems in companies.
The general bidding process method can be represented by the following diagrams
representing the two main situations of :
• knowledge re-use,
• and knowledge organization for re-use.
Technical solution designing with reuse of recurring items
During the bidding process different kinds of information are required to design
technical solutions. This information concerns the cost or has a technical aspect and
could come from previous bids. The bid process is resumed in the following figure :
Corporate memory
DBMS
Call for
proposal
Bid
Risk
Technical
solution
building
Proposal
costing
Technical solution
DBMS
To support the different kinds of user, i.e. bid managers and engineers the
developed prototype allows one to create new technical solutions by reusing previous
bids and the associated information concerning the products, the processes and the
resources or by adding new information that concerns the cost or the technical
feasibility.
After the bid step, the user has the possibility of organising the corporate memory
for possible future bids.
Organisation of the reuse : enterprise referential
The tool helps to organize corporate memory with a view to cost management
requirements orientation, capitalize and organize data as a management referential for
projects.
The corporate memory management is organised into three kinds of item :
• the temporary items
• the to-be examined items
• the recurring items.
The temporary items are an extraction of the information system. The user creates
completely new technical solutions, products, entities, processes or resources at the
moment of the bidding process. The temporary items form completely new technical
solutions.
From the temporary items, users have the possibility of creating items to be
examined. These items are those which could probably be reused for another bid.
This requires anticipating possible reusable items. It is worthwhile organising this
anticipation as it occurs in a non-urgent situation.
From the temporary and the to be examined items, recurring items are created.
Recurring items are those which are reused from previous bids and will certainly be
reused for future bids. Recurring items could be reused just as they are or brought up
to date. This data base represents the corporate memory. The user has the possibility
of anticipating in the long term. He will only select items which will certainly be
reused.
The designing of technical solutions with reuse of recurring items and the
organisation of this reuse can be resumed by the
following model base:
This is the generic model of the exchanges
Recurring
set :
Items
• exchanges inside the data base
• exchanges outside the database.
The tool promotes the re-use of the referential
Temporay
organized from previous bids while evaluating
Items
the particular context and the necessary
incorporation of new elements.
To be
It provides a means of managing the
Examined
exchanges inside the database.
Items
It is necessary to provide a means of managing
the exchanges outside the database. It is achieved
by the connection to the information system :
• drastic reduction of costs initiated at the best
step of the project
• the management of specialists’ interactions and trade-off during the bidding
process
• time-gaining for bid engineers (therefore more proposals can be done)
• standardisation of costing procedure
• generalisation of building a plan to cost related to a proposal.
The Risk Modelling
The decision about risk is mainly a trade off between global internal and external
risk exposures. External risk exposure -or tolerated risk level- is the independent
variable, internal risk -or incurred risk- is the dependent variable.
Information
System
Basket
The indicators are used at several phases of the decision process (bid/no bid,
subcontractors and partners choice, best and final technical solution choice, best and
final offer choice, and all along the process of choosing mitigation actions to reduce
risk). For example, technical solution criticality evaluation allows to mitigate risk to
reach an acceptable level of risk exposure, to compare different technical solutions, or
to take preventive or curative actions (insurance, provision, double providers...).
These factors are quantitative (financial or non-financial) or qualitative.
The Design phase approach uses classical risk assessment methods to estimate risk
exposure magnitudes (identification, scaling, ranking, prioritisation). Technical
solutions building and cost estimating are previous and necessary to risk estimating.
This phase takes into account risks as well as opportunities.
Risk exposure is calculated depending on risk categories: cost, schedule,
performance. The exposure can be finely expressed as a global exposure / cost. Risk
models allow the bid manager to develop alternatives, scenarios, and simulations of
mitigation actions impact.
Indicators are to be evaluated on the basis of AHP methods [14].
The Intelligence phase provides risk identification. This identification is made
through the RMCM content. Knowledge discovery tools are used, such as Case Based
Reasoning.
Such business level knowledge has to be built from scratch, using a specific
corporate management process and paradigms [1][7][13]. Among theses paradigms, a
risk ontology is defined as the storage support and the classification of risk
knowledge, specifically oriented towards management vision and business reuse.
Risk knowledge covers expert knowledge and historical data. The relative
importance of the two classes varies considerably, especially when one shifts from a
repetitive to a non-repetitive environment. The former requires systematic and
extensive use of data related to similar previous situations, even if experts are needed
to validate the information obtained. In the latter, since historical data are scarce by
definition, useful and subjective judgements are the main means of obtaining
information on the various factors that influence the bid process. The integration
between available historical data - unavoidably limited by uniqueness - and subjective
judgements elicited by specialists on the basis of previous experience of similar
situations is an inherent issue of the knowledge engineering process.
5
Conclusion
This paper analyses KDD methods and tools relevance to design and develop
ODSS.
In opposition with KDD methods and techniques which are data-pushed, the
adopted approach starts and is pulled by requirements engineering. This approach is
most common in the field of decision support systems design.
Requirements engineering is based on the analysis of both the decision context
inside the company (company requirements) and the decision maker activities
(cognitive process requirements). The result of the adopted approach is the design of
the knowledge necessary to decision-makers using the ODSS.
The presented ODSS is a very common need for strategic decision making within
process-organized firms, so these conclusions are of rather large application.
The PRIMA knowledge sources, which are in the RMCM, are at the following
level of the KDD process [5]:
Interpretation/
Evaluation
Datamining
Knowledge
Transformation
Preprocessing
Patterns
Selection
Preprocessed
data
Data
Target data
Transformed
data
RMCM
Fig. 3. Situation of the corporate memory of PRIMA (RMCM) in the KDD process
The question of deploying the different steps between the company existing
information systems and the RMCM is basic (for example to organise return on
experience from projects), and could use KDD methods. This problem is under study.
Nevertheless, as complexity and quantity of projects that would be over time
incorporated in the RMCM would lead to a profuse amount of data, KDD tools are
used: Case Based Reasoning (CBR) [8] to search the knowledge base, classification
and clustering techniques to help knowledge emergence.
But classification cannot be obtained by exploring raw data; it can be based only
on the use of the RMCM ontology as meta-knowledge. Clustering could lead
nevertheless to RMCM modifications, but at knowledge level and not metaknowledge level.
At business level, information aggregation is an absolute necessity. This
aggregation is the result of a co-operative process which builds inside the firm a
specific language and knowledge co-constructed by multiple decision-makers, as the
following figure shows:
RMCM Strategic decisions
Problem of Knowledge
Aggregation and
Communication
Co-constructed cooperative language
Company Information System
Fig. 4. RMCM knowledge sources regarding company's information system
4
References
1.
Alquier, A.M., Modélisation des systèmes d'information : modèle coopératif.
Application à l'audit et aux schémas directeurs d'informatique, Habilitation à diriger des
recherches, Université Toulouse 1, 1993
2.
Davenport T.H., Process innovation; Reengineering work through information
technology, Harvard Business School, 1993
3.
Desanctis G. & Gallupe R.B., A foundation for the study of group decision support
systems, Management Science, Vol. 33, No. 5, May 1987.
4.
Duncan W.R., A guide to the Project Management Body of Knowledge, PMI
Standards Committee, 1996
5.
Fayyad U.M., Piatetsky-Shapiro G. & Smyth P. " From Data Mining to Knowledge
Discovery: An Overview ", KDD 1995
6.
Holsapple C.W. & Whinston A.B., Decision support systems; a knowledge based
approach, West Publishing Company, 1996
7.
Kirn S. & O'Hare G., Cooperative knowledge processing, Springer Verlag, 1997
8.
Kolodner J., Case-Based Reasoning, Morgan Kaufmann, 1993
9.
Michaels J.V. & Wood W.P., Design to cost, Wiley, 1989
10. Kroenke D. & Hatch R., Management Information Systems, McGraw-Hill, 1994
11. Mintzberg H. G., The rise and fall of strategic planning, The Free Press, 1994
12. Newell A. & Simon H. A., Human problem solving, Prentice Hall, 1972
13. Nonaka I. &Takeuchi H., The knowledge creating company - how Japanese
companies create the dynamics of innovation, Oxford University Press, 1995
14. Saaty, T.L., The Analytical Hierarchy Process, McGraw-Hill, 1980
15. Schmidt K., Cooperative work: a conceptual framework, in "Distributed decision
making: cognitive models for cooperative work", Rasmussen J. Brehmer B., Leplat J. (ed.),
John Wiley & Sons Ltd 1991
16. Sprague R., & Carlson E., Building effective decision support systems, Prentice?Hall,
Inc., 1982