Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
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