Fuzzy C-Means Clustering For The Optimal Portfolio Of Machinery

Document Sample
Fuzzy C-Means Clustering For The Optimal Portfolio Of Machinery Powered By Docstoc
					           Fuzzy C-Means Clustering to
         Explore the Strategy Combination
               of Fuel Cell Industry

                   Hua-Kai Chiou, Gwo-Hshiung Tzeng
                  Benjamin J.C. Yuan & Chien-Pin Wang

                  National Chiao Tung University, Taiwan


IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA            1
 Agenda
     Introduction
     Multiple Criteria Decision Making Process
     Fuzzy C-Means Clustering
     Empirical Study
     Conclusions




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA   2
 Introduction

     According to the Protocol of United Nation Climatic
      Change Summit, there would be a 5% decrease of
      total emissions by 2012 against that of 1990, thus
      slowing the global warming process (Kyoto Protocol).

     Fuel cell is one type of electrical power device, but is
      unlike regular battery that is either thrown away after
      use or needing recharge.




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA              3
 …Introduction

     Fuel cell can be divided into Alkaline fuel cell (AFC),
      Phosphoric acid fuel cell (PAFC), Molten carbon fuel
      cell (MCFC), Solid oxide fuel cell (SOFC), Proton
      exchange membrane fuel cell (PEMFC), and Direct
      methanol fuel cell (DMFC).

     Fuel cell needs a type of fuel, hydrogen, to maintain
      its electrical power.




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA             4
 …Introduction

     Fuzzy AHP with multiple criteria decision
      analysis (MCDA) approach to find the preferred
      order of these development strategies.

     We further introduce fuzzy c-means to determine
      the optimal combination of these strategies.




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA        5
Table 1. Fuel Cell Technology Status, Research Institutes and Application Market




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA                               6
Table 2. Alternative Energy Development Policy from Major Countries (ITRI, 2001)

 Country      Projects/Actions              Alternative Energy/New Energy Policy

 Austria      Recycle energy power          50% of the financial aid for the equipment used on recycle energy in some
              generation project            areas.

 US           Technology development        US$2.4 billion to the original global warming foundation, increasing 40%
              foundation                    budget for developing the clean energy.

 Japan        Global warming action plan    Enhance technology development on fuel cell, solar energy and bio energy.
                                            Japan’s fuel cell development goal, 2.2 million watts by 2010.

 Germany      Revision on Power Parallel    Revision Act to target 10% of the recycle energy by 2010 and energy ratio
              Connection Bill               of 4%.

              Ecology tax act               Higher tax on petrochemical energy to reduce the total consumption of the
                                            petro fuel.

 Korea        1997 - 2006 energy            Plan to invest Korea Won 2,047 billion to develop energy protection
              technology development        technology, alternative energy, and clean energy. The alternative energy
              program                       includes fuel cell, solar energy, solar optic electronics and IGCC.

 Korea        New and recycle energy        Mainly to develop 12 alternative/ clean energy technologies, including fuel
              development act               cell. The incentives granted include 5% of the investment is tax free and up
                                            to 80% of the investment can enjoy the 5% low interest loan.


IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA                                                                         7
Table 3. 2000-2010 Market Forecasting on Global Fuel Cell Applications (Hester,2001)


                                     2005            2010              2000~2010
           Applications
                                  (US$ billion)   (US$ billion)   Compounded Growth Rate

        Power Generation               3.8            10.5               26.64%


          Vehicle Power                1.8            3.9                20.61%


         Portable Power                1.2            4.5                35.6%


             Others                    1.7            4.4                21.78%


              Total                    8.5            23.3




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA                                            8
 Multiple Criteria Decision Making Process

     Identifying the nature of problems;
     Defining the decision variables;
     Determining the appropriate evaluation method;
     Analyzing and evaluating step by step;
     Proposing results and conclusion for decision
      making.




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA        9
 Identifying the nature of problems
     MCDA problems:
          What are the aspects, considered criteria, sub-criteria …?
          Qualitative or quantitative?
          Crisp, fuzzy, grey or rough set data?
          Does it exists threshold value of criteria?
          Maximum, minimum or anticipated value seeking for criteria?

     MOP problems:
          What are the goal, constraints and right-hand sight values?
          Qualitative or quantitative?
          Crisp, fuzzy, grey or rough set data?
          Single-objective, bi-objectives or multi-objectives problem?
          Single-level, multi-level, multi-stage programming?

IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA                       10
 Determining the evaluation method

    MCDA approaches:
         AHP;
         SAW;
         Grey Relation Analysis;
         TOPSIS;
         VIKOR;
         ELECTRE;
         PROMETHEE;
         Statistics Inferring

IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA   11
 …Determining the evaluation method

     MOP techniques:
          Goal Programming;
          DEA;
          De Novo;
          TOPSIS;
          Compromise Solution;
          εconstraints;
          Game Theory.



IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA   12
 Fuzzy C-Means Clustering

     It is a branch in multivariate analysis and an
      unsupervised learning in pattern recognition.
     Hard C-Mean Clustering (HCM, K-Means)
     The objective function of HCM can be defined as
      following:


                                                   (1)




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA     13
 …Fuzzy C-Means Clustering

     The objective function of Fuzzy C-Means
      clustering can be defined as following:

                                                  (2)




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA   14
 …Fuzzy C-Means Clustering
       The FCM algorithm can be summarized as
        following steps:
     1). Choose an initial partition membership matrix U0;
     2). Set the stop condition as follows:

                                                                       (3)

     3). While not stop at condition (3) do
          3-1). Compute reference vectors for each part family using
                Eq.(4)

                                                                       (4)

IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA                        15
 …Fuzzy C-Means Clustering

           3-2). Update the part membership matrix according to (5)



                                                                      (5)


           3-3). Evaluate Jm(U,V) using (6).

                                                                      (6)




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA                       16
 Empirical Study

     This research is to develop and evaluate the
      proposed strategies of Fuel Cell Industry utilizing
      Fuzzy AHP with MCDM approach.

     Fuzzy C-Means Clustering to find the optimal
      combination of strategies for resources limitation.




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA        17
                     Fig. 1 Hierarchical frame of evaluation model
IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA                      18
          Table 1. Defuzzified Fuzzy Weights of Evaluated Criteria




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA                      19
           Table 2. Defuzzified Performance Score of Evaluated Data




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA                       20
       Table 3. Defuzzified Synthetic Values of Development Strategies




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA                          21
     From Table 3, we can conclude the preferred
      order of proposed development strategies of fuel
      cell industry as follows:




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA      22
      The first five development strategies to be
       implemented suggested by experts who
       participated this research are
          1. Policy or Regulations for Clean Energy;
          2. Increase R&D Budget and Transfer Technology
             from Overseas;
          3. Establish National Program and enhance
             coordination and promotion mechanism;
          4. Carefully Select the Niche Products and its
             applications; and
          5. Develop Demonstrated Zone or System with
             Financial Support.

IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA            23
  Table 4. Grade of Membership of Strategies for Fuel Cell Industry (three-cluster case)




  Table 5. Grade of Membership of Strategies for Fuel Cell Industry (four-cluster case)




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA                                           24
     Table 4 shows the three-cluster combination:
         The first cluster includes two strategic combinations,
          that is, Low interest loan (S6) and Train marketing &
          planning professionals (S11).
         The second cluster includes four strategic
          combinations, that is, Increase R&D budget and
          transfer technology from overseas (S1), Carefully
          select the niche products and its applications (S2),
          Establish national program and enhance coordination
          and promotion mechanism (S8), and Policy or
          regulations for clean energy (S9).


IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA                    25
         The third cluster includes seven strategic
          combinations, that is, Financial aid to new users (S3),
          Sponsor to build infrastructure and tax aid (S4), Tax
          incentives and other incentive measures (S5), Make
          the product standard and establish certified testing
          center (S7), Re-enforce education training promotion
          and demo (S10), Efficiently manage business
          strategies (S12), and Develop exemplified zone or
          system with financial support (S13).




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA                 26
     Table 5 presents the strategy combination of
      four-cluster cases;
         The strategic combinations of the first two clusters
          have the same contents. It only split the third cluster to
          two clusters;
         The third cluster includes seven strategic
          combinations, that is, Financial aid to new users (S3),
          Sponsor to build infrastructure and tax aid (S4), Tax
          incentives and other incentive measures (S5), and
          Develop exemplified zone or system with financial
          support (S13).

IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA                   27
         The fourth new cluster includes three strategic
          combinations, that is, Make the product standard and
          establish certified testing center (S7), Re-enforce
          education training promotion and demo (S10), and
          Efficiently manage business strategies (S12).




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA                  28
     We further explore the results of different fuzzy
      classification with strategic preferred order
      based on individual synthetic value, we can
      easily find that,
         Both strategies in the first cluster (S6 and S11) have
          the lowest preference.
         All of the strategies in the second cluster (S1, S2, S8
          and S9) have the higher synthetic values.
         In addition, the strategies in third or fourth cluster have
          the medium synthetic values.


IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA                    29
 Conclusions

     Facing the dynamic change of global business
      environment, although the evaluators give
      different weights on criteria, from group decision
      evidence, the first five important criteria are
      technology enhancement (0.1111), R&D
      investment (0.0935), explicit government policy
      (0.0760), reducing production cost (0.0759), and
      R&D manpower (0.0692).



IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA        30
 …Conclusions

     It indicates that technology enhancement and
      R&D investment will definitely influence the
      fulfillment in such emerging industry. For the
      role of government, how to make an explicit
      policy is very important especially in such
      emerging technology and industry. In addition,
      most of the participated experts agree that how
      to reduce production cost and how to recruit and
      manage the R&D manpower also are the critical
      factors for getting into such new applied field.
IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA      31
 …Conclusions

     This research has successfully demonstrated
      the appropriateness of fuzzy c-means clustering
      for solving the optimal strategic combinations for
      fuel cell industry in Taiwan. The widely used
      technique, AHP, fuzzy c-means clustering can
      provide some useful information for coping with
      real MCDA problems.




IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA        32
                              THE END

          Thanks for your patience
          and attention to listening.

                   E-mail: ghtzeng@cc.nctu.edu.tw (G.H. Tzeng)
                          hkchiou@ebtnet.net (H.K. Chiou)

IFORS. July 11 - 15, 2005. Honolulu Hawaii, USA                  33

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:10
posted:8/6/2011
language:English
pages:33