18 pages

Fuzzy cognitive maps as a back end to knowledge-based systems in geographically dispersed financial organizations

of 18
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
This paper addresses the problem of designing a knowledge management methodology tool to act as a decision support mechanism for geographically dispersed financial enterprises. The underlying research addresses the problem of information capture and
  & Research Article  Fuzzy Cognitive Maps as a BackEnd to Knowledge-based Systemsin Geographically DispersedFinancial Organizations George Xirogiannis 1 *, Michael Glykas 1 and Christos Staikouras 2 1 Department of Financial and Management Engineering, University of the Aegean, Greece 2 Department of Accounting and Finance, Athens University of Economics and Business, Greece This paper addresses the problem of designing a knowledge management methodology tool toact as a decision support mechanism for geographically dispersed financial enterprises. Theunderlying research addresses the problem of information capture and representation in finan-cial institutions in order to provide an implementation of the virtuous cycle of knowledge flow.The proposed methodology tool utilizes the fuzzy causal characteristics of Fuzzy CognitiveMaps (FCMs) to generate a hierarchical and dynamic network of interconnected financial per-formance concepts. By using FCMs, the proposed mechanism simulates the operational effi-ciency of distributed organizational models with imprecise relationships and quantifies theimpact of the geographically dispersed activities to the overall business model. Generic adap-tive maps that supplement the decision-making process present a roadmap for integrating hier-archical FCMs into the business model of typical financial sector enterprises. Copyright # 2004John Wiley & Sons, Ltd. INTRODUCTION Knowledge has been defined by Western philoso-phy (Russell, 1989) as ‘justified true belief’. How-ever, as long as there is a chance that this belief ismistaken and as long as there is an evolution of technologies, theories, practice and behaviours,this definition invites individuals and groups todevelop constantly ‘what they think that theyknow’ (Nonaka and Takeushi, 1994). This continu-ous process of creation of new insights and beliefsis what fuels the entire paradigm of knowledgemanagement and even constitutes the fundamentalrationale for the existence of an enterprise. Someargue that instead of merely solving problems,organizations create and define problems, developand apply new knowledge to solve problems andthen develop further new knowledge through theaction of problem solving. In this view an enter-prise creates continuously new knowledge throughaction and interaction, not acting simply as aninformation-processing machine (Nonaka andTakeushi, 1994; Spencer and Grant, 1996).Currently, there is a pressing need for sharingknowledge among the financial managers (e.g.CFO, Head of Accounting, Head of Budgeting,Head of Business Planning). The effect of knowl-edge attrition for customers and investors is knowl-edge intensive. Financial enterprises lose (onaverage) half of their knowledge base every 5–10years due to the turnover of their employees. Thefact is that financial managers are usually morecommitted to their specialized profession, thus Knowledge and Process Management Volume 11 Number 2 pp 137–154 (2004)Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/kpm.199 Copyright # 2004 John Wiley & Sons, Ltd. *Correspondence to: George Xirogiannis, Department of Finan-cial and Management Engineering, University of the Aegean,31 Fostini Street, Chios, 82 100, Greece.E-mail: g.xirogiannis@fme.aegean.gr  losing sight of the enterprise as a whole. Withouteffective sharing and maintenance of knowledge,financial institutions risk losing their crucial knowl-edge through labour mobility. Moreover, knowl-edge has the idiosyncrasy that its value can only be realized when it is put in use (i.e. no shelf value).The more knowledge put in use, the more its valueis appreciated (e.g. efficient use through theincrease in confidence of application and learning),rather than depreciated, as other physical assets of machinery or natural resources.This paper addresses the problem of designing anovel knowledge management (KM) methodologytool to act as a decision support mechanism forgeographically dispersed financial organizations(e.g. multi-national enterprises, multi-branch banks). The proposed methodology tool utilizesthe fuzzy causal characteristics of fuzzy cognitivemaps (FCMs) to generate a hierarchical anddynamic network of interconnected financial per-formance concepts. By using FCMs, the proposedmechanism simulates the operational efficiency of distributed organizational models with impreciserelationships and quantifies the impact of the geo-graphically dispersed activities to the overall busi-ness model.From the KM perspective, it is the belief of thispaper that optimal decision-making can beachieved through the creation of a virtuous circleof knowledge flow: creating and adding knowl-edge, successful searching for the required piecesof knowledge, feedback for impetus to assert qual-ity knowledge again.Primarily, the proposed model targets the princi-pal beneficiaries and stakeholders of KM projects(financial administration, change management lea-ders, etc.), assisting them to reason effectivelyabout the status of financial performance metrics,given the (actual or hypothetical) implementationof a set of business model changes. Nevertheless,the explanatory nature of the mechanism can proveto be useful in a wider educational setting.This paper consists of seven sections. The nextsection presents a short literature overview, whilethethirdsectionaddressesthecontemporaryknow-ledge capture and representation problem of finan-cial managers and justifies the need for a noveladaptive knowledge management methodologytool. The fourth section presents an overview of the proposed system, the fifth section discussesthe new approach to knowledge modelling basedon FCMs and the sixth section discusses the majoradvantages of the proposed tool. The final sectionconcludes this paper and briefly discusses futureresearch activities. LITERATURE OVERVIEW FCMs as a modelling technique FCMs are a modelling methodology for complexdecision systems, which srcinated from the combi-nation of fuzzy logic (Zadeh, 1965) and neural net-works. An FCM describes the behaviour of asystem in terms of concepts; each concept repre-sents an entity, a state, a variable or a characteristicof the system (Dickerson and Kosko, 1997).FCM nodes are named by such concepts formingthe set of concepts  C ¼ { C 1 ,  C 2 , . . . , C n }. Arcs ( C  j ,  C i )are oriented and represent causal links betweenconcepts; that is how concept  C  j  causes concept  C i .Arcs are elements of the set  A ¼ {( C  j , C i )  ji }  C  C .Weights of arcs are associated with a weight valuematrix W n  n , where each element of the matrix w  ji 2 [  1, . . . ,1]  R  such that if ( C  j ,  C i ) = 2  A  then w  ji ¼ 0 else excitation (respectively inhibition)causal link from concept  C  j  to concept  C i  gives w  ji > 0 (respectively  w  ji < 0). The proposed metho-dology framework assumes that [  1, . . . ,1] is afuzzy bipolar interval, bipolarity being used as ameans of representing a positive or negativerelationship between two concepts.In practice, the graphical illustration of an FCMis a signed graph with feedback, consisting of nodes and weighted interconnections (e.g.  ! Weight ).Signed and weighted arcs (elements of the set  A )connect various nodes (elements of the set  C ) repre-senting the causal relationships that exist amongconcepts. This graphical representation (e.g.Figure 1) illustrates different aspects in the beha-viour of the system, showing its dynamics (Kosko,1986) and allowing systematic causal propagation(e.g. forward and backward chaining).  Figure 1 Basic constructs of FCMs RESEARCH ARTICLE Knowledge and Process Management  138 G. Xirogiannis  et al.  Positive or negative sign and fuzzy weightsmodel the expert knowledge of the causal relation-ships (Kosko, 1991). Concept  C  j  causally increases C i  if the weight value  w  ji > 0 and causally decreases C i  if   w  ji < 0. When  w  ji ¼ 0, concept  C  j  has no causaleffect on  C i . The sign of   w  ji  indicates whetherthe relationship between concepts is positive( C  j ! w  ji C i ) or negative ( C  j ! w  ji   C i ), while the valueof   w  ji  indicates how strongly concept  C  j  influencesconcept  C i . The forward or backward direction of causality indicates whether concept  C  j  causes con-cept  C i  or vice versa (e.g. Figure 2).Simple variations of FCMs mostly used in busi-ness decision-making applications may take triva-lent weight values [  1,0,1]. This paper allowsFMCs to utilize fuzzy word weights like strong,medium or weak, each of these words being a fuz-zy set to provide complicated FCMs. In contrast,Kwahk and Kim (1999) adopted only a simplerelative weight representation in the interval[  1, . . . ,1]. To this extent, Kwahk and Kim (1999)offered reduced functionality since it does notallow fuzzy weight definitions.Generally speaking, FCM concept activations taketheir value in an activation value set  V  ¼ {0,1} or{  1,0,1} if in crisp mode or [    ,1] with    ¼ 0 or 1 if in fuzzy mode. The proposed methodologyframework assumes fuzzy mode with    ¼ 1. At step t 2 N  , each concept  C  j  is associated with an inner acti-vation value  a  jt 2 V  , and an external activation value e a  j t 2 R . FCM is a dynamic system. Initialization is a  j 0 ¼ 0. The dynamic obeys a general recurrent rela-tion  a t þ 1 ¼  f   (  g ( e at ,  W  T  a t )),  8 t    0, involving weightmatrix product with inner activation, fuzzy logicaloperators (  g ) between this result and external forcedactivation and finally normalization (  f  ). However,this paper assumes no external activation (hence nofuzzy logical operators), resulting in the followingtypical formula for calculating the values of conceptsof FCM: a t þ 1 i  ¼  f  X n j ¼ 1 ;  j 6¼ i w  ji a t j 0@1A  ð 1 Þ where  a it þ 1 is the value of concept  C i  at step  t þ 1, a  ja the value of the interconnected concept  C  j  atstep  t ,  w  ji  is the weighted arc from  C  j  to  C i  andf: R ! V   is a threshold function, which normalizesactivations. Two normalization functions are us-ually used. The unipolar sigmoid function where > 0 determines the steepness of the continuousfunction  f  ð x Þ ¼  11 þ e   x . When concepts can be nega-tive ( < 0), function  f  ð x Þ ¼  tanh ð x Þ  can also beused.To understand better the analogy between thesign of the weight and the positive/negative rela-tionship, it may be necessary to revisit the charac-teristics of the fuzzy relation (Kaufmann, 1975; Lee et al. , 2002). A fuzzy relation from a set  A  to a set  B or (  A,B ) represents its degree of membership inthe unit interval [0, 1]. Generally speaking, sets  A and  B  can be fuzzy sets. The corresponding fuzzymembership function is  m f:  A  B ! [0, 1]. There-fore,  m  f  (x, y) is interpreted as the ‘strength’ of thefuzzy membership of the fuzzy relation ( x,y )where  x 2  A  and  y 2 B . Then this fuzzy relation con-cept can be denoted equivalently as  x !   f   y  andapplied to interpret the causality value of FCM,since  w  ji  (the causality value of the arc from nodes C  j  to  C i ) in a certain FCM is interpreted as thedegree of fuzzy relationship between twonodes  C  j  and  C i . Hence,  w  ji  in FCMs is the fuzzymembership value  m  f  ( C  j , C i ) and can be denotedas  C  j ! w  ji C i .However, we understand that the fuzzy relation(weight) between concept nodes is more generalthan the srcinal fuzzy relation concept. This is because it can include negative (  ) fuzzy relations.Fuzzy relations mean fuzzy causality; causality canhave a negative sign. In FCMs, the negative fuzzyrelation (or causality) between two concept nodesis the degree of a relation with a ‘negation’ of aconcept node. For example, if the negation of aconcept node  C i  is noted as   C i , then  m  f  ( C  j , C i ) ¼ 0.6 means that  m  f  ( C  j ,   C i ) ¼ 0.6. Conversely, m  f  ( C  j , C i ) ¼ 0.6 means that  m  f  ( C  j ,   C i ) ¼ 0.6.FCMs help to predict the evolution of the system(simulation of behaviour) and can be equippedwith capacities of Hebbian learning (Kosko,1986a, 1998). FCMs are used to represent and to model the knowledge on the examining system.Existing knowledge of the behaviour of the systemis stored in the structure of nodes and interconnec-tions of the map. The fundamental difference between FCMs and neural networks is in the factthat all the nodes of the FCM graph have a strongsemantic defined by the modelling of the concept,whereas the nor input/nor output nodes of thegraph of the neural network have a weak semantic,only defined by mathematical relations.  Figure 2 Simple FCM Knowledge and Process Management RESEARCH ARTICLE  Fuzzy Cognitive Maps 139  Applications of fuzzy cognitive maps Over the last 10 years, a variety of FCMs have beenused for capturing—representing knowledge andintelligent information in engineering applications,for instance, geographical information systems Liuand Satur (1999) and fault detection (Ndouse andOkuda, 1996; Pelaez and Bowles, 1995). FCMshave been used in modelling the supervision of dis-tributed systems (Stylios  et al. , 1997). FCMs havealso been used in operations research (Craiger et al. , 1996), web data mining (Hong andHan, 2002; Lee  et al. , 2000), as a back end tocomputer-based models and medical diagnosis(e.g. Georgopoulos  et al. , 2002).Several research reports applying basic conceptsof FCMs have also been presented in the field of  business and other social sciences. Research inAxelrod (1976) and Perusich (1996) have usedFCM for representing tacit knowledge in politicaland social analysis. FCMs have been successfullyapplied to various fields such as decision makingin complex war games (Klein and Cooper, 1982),strategic planning (Diffenbach, 1982; Ramaprasadand Poon, 1985), information retrieval ( Johnsonand Briggs, 1994) and distributed decision processmodelling (Zhang  et al. , 1994). Research like that of Lee and Kim (1997) has successfully applied FCMsto infer rich implications from stock market analy-sis results. Research like that of Lee and Kim (1998)also suggested a new concept of fuzzy causal rela-tions found in FCMs and applied it to analyse andpredict stock market trends. The inference power of FCMs has also been adopted to analyse the compe-tition between two companies, which are assumedto use differential games mechanisms to set uptheir own strategic planning (Lee and Kwon,1998). FCMs have been integrated with case-basedreasoning technique to build organizational mem-ory in the field of knowledge management (Noh et al. , 2000). Recent research adopted FCMs to sup-port the core activities of highly technical functionslike urban design (Xirogiannis, 2004). Summariz-ing, FCMs can contribute to the construction of more intelligent systems, since the more intelligenta system becomes, the more symbolic and fuzzyrepresentations it utilizes.In addition, a few modifications have been pro-posed. For example, the research in Silva (1995)proposed new forms of combined matrices forFCMs, the research in Hagiwara (1992) extendedFCMs by permitting non-linear and time delay onthe arcs, and the research in Schneider  et al. (1995) presented a method for automatically con-structing FCMs. More recently, Liu and Satur(1999) have carried extensive research on FCMs,investigating inference properties of FCMs, pro-posed contextual FCMs based on the object-oriented paradigm of decision support and appliedcontextual FCMs to geographical information sys-tems (Liu, 2000). KNOWLEDGE MANAGEMENT INFINANCIAL ENTERPRISES The decision-making problem Financial management is one of the major areas of  business of financial sector institutions. The risksundertaken by fund managers in their investmentdecisions affect directly the institutions’ businesscontinuity, profitability and reputation. Fund man-agers’ work allows direct performance measure-ment, while the rewards based on measuredperformance affect their career prospects and jobsecurity. Junior financial managers are thereforemore risk averse in their portfolio managementdecisions, which they ‘herd’ to follow the markettrends in avoiding investment risks above the mar-ket objective level, whereas senior managers aregiven more discretion in their investment deci-sions. The high monetary rewards, considerationsof job security and market-relative measurementof performance induce financial managers’ self-interests and ownership of critical knowledge,which may create effective investment decisions.Therefore creating an open environment to shareanalytical skills of market signals, intimate knowl-edge of clients, or capability to sense political atmo-sphere and market sentiment among financialmanagers can minimize any miscalculated risks. The need for knowledge sharing Knowledge can be reused potentially an infinitenumber of times without wearing off or needingrepair. With each subsequent application of infor-mation processing, the experience of use and back-ground understanding of environmental settings builds up. This potential to apply knowledge aninfinite number of times allows for economies of scale, which may reduce radically the transactionand/or operational costs. Hence the potential valueof knowledge can only be fully realized if it is lever-aged effectively through sharing and reuse.Intimate, contextual and firm-specific knowledgeis an intangible asset that is costly to develop,hard to replace and time consuming to replicate.Possession of intimate knowledge with real valueto customers allows enterprises to differentiatethemselves from competitors. Coordinated RESEARCH ARTICLE Knowledge and Process Management  140 G. Xirogiannis  et al.  processing of disparate pieces of information todevelop skills and capabilities in a way that helpsan enterprise to achieve its goals is a form of intel-lectual capital and a valuable source of competitiveadvantage. Individuals with different culture andexperiences lack the contextual links associatedwith the enterprise’s specific knowledge. Commu-nication of knowledge separated from its contexthinders thorough understanding and detachesintuitive interpretation through symbolic environ-mental associations. Knowledge so communicatedis inevitably abstract. It consists of generalized ana-lyses of events, and the recipient may find it diffi-cult to comprehend its relevance to a specificsituation. Indeed, skilful masters differentiatefrom novices by the intimate understanding of the tasks and the environment linked togetherwith specific contextual interacting factors that dif-ferentiate skillful masters from novice beginners.Sharing, and only communication, of knowledgecarries a common context where individuals caninteract and engage in a mutual dialogue with com-mon understanding. Effective sharing creates asynergy with value where the sum is more thanits parts. Therefore, effective dissemination of spe-cialized contextual knowledge among members of the firm minimizes miscalculation or mispercep-tion of risks. KM and transaction costs reduction Sharing of knowledge involves information flows between the distributed organizational entities of an enterprise. The cost—benefit analysis of knowl-edge flow must consider the enterprise’s specificenvironmental factors in terms of co-location of decision rights and knowledge. Colocation refersto the delegation of decision-making authority tothe party holding necessary knowledge (e.g. discre-tionary investment decisions made by financialmanagers within institutional guidelines).The cost of no knowledge sharing can be propor-tional to the decision-making costs (in practice, thesum of agency and knowledge transfer costs).Agency cost arises from self-interested individualsin conflict with management when decisions aredelegated to them. The time and effort spent bymanagement and possibly the need for policyadjustments, remuneration packages and coordina-tion of individual members are all agency costs.The knowledge transfer costs arise from training,practice and experience needed to equip indivi-duals with the necessary knowledge.The distance of decision rights from the CEOoffice affects the total decision-making costs andcauses trade-offs between suboptimal decisionsand possession of necessary knowledge. The lackof embedded know-how at central managementcauses suboptimal decisions to be made when thedecision-maker is away from sources of knowl-edge; therefore the knowledge transfer costs owingto poor information sharing is very high. On theother hand, when some knowledge carrier awayfrom central management makes a decision, theneed for control and coordination to maintain con-sistency and integrity incurs high agency costs.There is a cost-optimal point where the sum of the agency and transfer costs is at its minimum.Effective knowledge sharing lowers the costs byshifting the optimal point of total organizationalcosts downward, i.e. shifting the knowledge trans-fer cost curve downward. This component identi-fies the agency and knowledge transfer costsinvolved in sharing knowledge in the currentenvironment setting to ensure organizational fit. FUNDAMENTALS OF THEMETHODOLOGY TOOL The knowledge flow cycle Many financial analysts and investigation reportsabout multi-million dollar losses at Sumitomo andBarings attribute the losses to the neglect of payingattention to operational risk. Operational riskresults in losses due to deficiencies in information,personnel unavailability, human error, and inade-quate procedures and control. Also, financial man-agers who have practical knowledge about tradingrisks become reluctant to share their knowledgewith other organizational entities. By nature, mostof these risks are difficult to quantify and thereforeare handled arbitrarily. Thus, many analystsrecommend that financial enterprises utilize com-prehensive systems for capturing and monitoringrisk, accompanied by an automated risk-reportingprocess. These reports should be easily read andunderstood by top management.Such problems require technology and manage-ment tools to motivate financial managers’ willing-ness to offer information, identify the need forinformation processing, support top managementto deal with multi-attribute problems and achieve balanced and well-informed decisions. It is theview of this paper that optimal decision-makingcan be achieved through the creation of a virtuouscycle of knowledge flow (Figure 3). This paperadopts this knowledge flow sequence to propose,under the non co-location situation, an adaptiveenabling mechanism that will allow a financialenterprise to integrate knowledge capture as a back end to its intelligent decision-making process. Knowledge and Process Management RESEARCH ARTICLE  Fuzzy Cognitive Maps 141
Related Documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!