Docstoc

슬라이드 제목 없음

Document Sample
슬라이드 제목 없음 Powered By Docstoc
					Artificial Intelligence in
 Electronic Commerce
              - PRICAI 2000 Conference -



                   August 30, 2000



                     Jae Kyu Lee

            Graduate School of Management
Korea Advanced Institute of Science and Technology, and
      International Center for Electronic Commerce
                 (jklee@msd.kaist.ac.kr)


                                                          1
 What is Electronic Commerce?

   Platform: Internet, Web
   Applications
     • Advertisement
     • Online Sales: e-Marketplaces
     • Online Purchase: Comparison Shopping
     • eProcurement Systems for B2B EC
     • Customer Relationship Management (CRM)
     • Supply Chain Management (SCM)
     • Protocol: ebXML



Jae Kyu Lee                              PRICAI2000   2
 What AI components are used
           for EC?
      Expert System / Knowledge-based System-Rules,
       Frames(Objects), Inference, CBR, Ontology
      Machine Learning and Knowledge Acquisition
       Systems
      Agents: ACL and Multi-Agent Systems
      Data Mining and Web Mining
      Voice Recognition
      Image Processing
      Natural Language Processings
      Machine Translation




Jae Kyu Lee                                  PRICAI2000   3
          Table of Contents

   From XML to XRML
   eCatalog and Directory Balancing Theory
   Configuration with Options
   Agents Based Commerce Architecture
   Ontology Initiatives in EC
   From Data Mining to Web Mining



Jae Kyu Lee                       PRICAI2000   4
     Challenge From Web
                            ES/KBS                Web/KMS/XML

 System Title     Knowledge Based         Knowledge Management
                  System                  System

 Knowledge        Rule, Logic, Frames     HTML, XML, Hypermedia,
 Representation   [Structured]            VRML [Semi-structured]

 Inference        Chaining,               HyperLink,
                  Approximate Reasoning   Search
 Knowledge        System                  Human
 Processor
 Scope of         Specific Knowledge      Global Knowledge Portal
 Knowledge        Base
 Tools            Domain Specific         General Purpose

 Popularity       Limited                 Everybody, Every day




Jae Kyu Lee                                            PRICAI2000   5
      Next Step of Web
Web Page
                                Web
                              Browsing
       HTML                                      Human


                                  Data
        XML                                      Agent
                               Processing



                                                 • Agents
        XRML                                     • Forms
                                 Rule
                              Processing



               eXtensible Rule Markup Language

Jae Kyu Lee                                  PRICAI2000     6
  Illustrative XML form and
DTD(Document Type Definition)
 XML for form

 <?XML version=“1.0”?>
 <form>purchase_requisition_form
 <requisition_section>ICEC</requisition_section>
    <item>Asset_Purchasing</item> DTD for XML form
    <account>P-1234</account>
    <product_description>             <!ELEMENT form (requisition_section, item,
      <product_type> PC </product_type>           account, product_description,
      <product_name>SPC-1111</product_name> source_or_purchase, purchase_method)>
                                      <!ELEMENT requisition_section(#PCDATA)>
      <estimated_price>1000</estimated_price>
      <quantity>1</quantity>          <!ELEMENT product_description(product_type,
      <specification>                             product_name, estimated_price, quantity,
                                                 specification)>
         <processor_type>pentium_III</processor_type>
         <hard_size>10GB</hard_size>  <!ELEMENT product_type (#PCDATA)>
                                      <!ELEMENT specification (processor_type,
         <main_memory>64MB</main_memory>
         <monitor_size>17inch</monitor_size>      hard_size, main_memory, monitor_size)>
      </specification>                <!ELEMENT processor_type (#PCDATA)>
   </product_ description>
   <source_of_purchase>domestic</ source_of_purchase>
   <purchase_method>single_estimation</purchase_method>
 </form>


Jae Kyu Lee                                                                PRICAI2000        7
              eXtensible Stylesheet
                Language (XSL)

     To display it on web browser
        • XML should be translated to HTML via eXtensible
          Stylesheet Language (XSL).

     Translation Architecture

       Valid XML                              Well-formed XML

              XML        XSL        HTML
                                                      DTD
               DTD                             Web browser




Jae Kyu Lee                                   PRICAI2000     8
 Research from AI Community:
     Data/Objects in XML
   Case Based Markup Language (CBML)
     - Cases in XML

   Artificial Intelligence Markup Language (AIML)
    - XML in a Chat-bot

   Multi-Modal Dialog Language (MMDL)
    - XML for Agents Communication, Online Stock Trading

   A*E XML Parser [PRICAI 2000, Workshop]
    - XML  Java

   We need Ontology Study



Jae Kyu Lee                                   PRICAI2000   9
eXensible Rule Markup Language

   Rule Structure Language
     • Representation of rules in Markup Syntax


   Rule Triggering Language
     • Trigger an inference engine against the rule
       base


   Rule Identification Language
     • Representation of HTML including the
       variables used in RSL
     • Identification of relevant rules to a paragraph



Jae Kyu Lee                                PRICAI2000   10
                Architecture of XRML
                                              Expert System Agent

                                                                Inference
                                          Rule Base   Call
                                                                 Engine
                           Transform

                                                                      Trigger
              Knowledge                                               Inquiry
                                                                                        Answer
                             RSL
Knowledge       Editor
 Engineer
                                                                  RTL in
                                Consistency                       Forms

                XML/
                           XML/RIL                           Object-type Forms
                HTML



                                                             Workflow


              Web Brower


            Jae Kyu Lee                                                    PRICAI2000     11
                  Display for Human’s
                    Comprehension

.   Research budgets is organized within limits of the contract deposit and
    distributed to each accounts. The use of research budget is as follows:

        If the budgetary source is the type-P research fund, the spendable
        items are student salary and expenses for data collection

    The use of research budget could be changed by the decision of board of
    directors. Only a person who possess the right of modification can
    change the regulation and the changed regulation is enforced after the
    president’s approval.

        If the residual of budgetary source less than purchase amount, then
        purchase is halted


    Jae Kyu Lee                                             PRICAI2000       12
                   Illustrative HTML/RIL
             Research budgets is organized within limits of the contract deposit and distributed
             to each accounts. The use of research budget is as follows:
 Browsed           If the budgetary source is the type-P research fund, the spendable items are
 Display           student salary and expenses for data collection.

         <HTML>
         <p> Research budgets is organized within limits of the contract deposit and
         distributed to each accounts. The use of research budget is as follows: </p>
         <RIL>
HTML/RIL <RILHeader>
         <RILTitle>budgetary appropriateness</RILTitle>
         </RILHeader>
         <RILBody>
          If the <var>budgetary source</var> is the <val>type-P research fund</val>, the
         spendable <var>items</var> are<val>student salary</val>and <val>expenses for
         data collection</val>
         </RILBody>
         </RIL>
         <HTML>


     Jae Kyu Lee                                                          PRICAI2000       13
Rules and Rule Structure Language

               Rule Group: Budgetary Type Constraints

                Rule Title: Budgetary Appropriateness
                  IF          (budgetary-source IS type-P research fund)
Rule                AND        ((item IS student’s-salary
                          OR (item IS data-collection-expenses))
                  THEN          permission IS allowed


                <RSL>
                <RSLRulegroup>Budgetary Type Constraints</RSLRulegroup>
                <RSLRule>
                <RSLRuleTitle>budgetary_appropriateness</RSLRuleTitle>
RSL
                <RSLRuleBbody>
                <IF> (AND (<var>budgetary_source</var> IS <val>type-P research fund</val>)
                        (OR   (<var>item</var> IS <val>student’s_salary</val>)
                              (<var>item</var> IS <val>data_collection_expenses</val>)))</IF>
                <THEN> <var>permission</var> IS <val>allowed</val> </THEN>
                </RSLRulrBody>
                </RSLRule>
                </RSL>


Jae Kyu Lee                                                                           PRICAI2000   14
      Consistency between RIL and RSL

         <HTML>
RIL      The research fund is …………..           ……….
         <RIL>
         <RILHeader>
         <RILTitle>budgetary appropriateness</RILTitle>
         </RILHeader>
         <RILBody>
          If the <var>budgetary source</var> is the <val>type-P research fund</val>, the
         spendable <var>items</var> are<val>student salary</val>and <val>expenses for data
         collection</val></RILBody></RIL><HTML>



         <RSL>
RSL      <RSLRulegroupTitle>Budgetary Type Constraints</RSLRulegroupTitle>
         <RSLRule>
         <RSLRuleTitle>budgetary appropriateness</RSLRuleTitle>
         <RSLRuleBbody>
         <IF> (AND (<var>budgetary_source</var> IS <val>type-P research fund</val>)
               (OR (<var>item</var> IS <val>student’s salary</val>)
                      (<var>item</var> IS <val>data collection expenses</val>)))</IF>
         <THEN> <var>permission</var> IS <val>allowed</val> </THEN>
         </RSLRulrBody></RSLRule></RSL>


 Jae Kyu Lee                                                            PRICAI2000           15
                Architecture of XRML
                                              Expert System Agent

                                                      Call      Inference
                                          Rule Base              Engine
                           Transform

                                                                      Trigger
              Knowledge                                               Inquiry
                                                                                        Answer
                             RSL
Knowledge       Editor
 Engineer
                                                                  RTL in
                                Consistency                       Forms

                XML/
                           XML/RIL                           Object-type Forms
                HTML


                                                               Workflow


              Web Brower


            Jae Kyu Lee                                                    PRICAI2000     16
          Form as an Object
{{requisition-2000/6/28-1                     {{product_description
     instance_of: purchase_requisition_form       part_of: purchase_requisition_form
     requisition_section: KGSM                    product_type: PC
     item: asset_purchasing                       product_name: SPC-111
     account: P-1234                             estimated_price: 1000
     product_description:                         quantity:1
     source_of_purchase: domestic                 specification:      }}
    purchase_method: single_estimation }}

                                              {{specification
                                                 part_of: product_description
                                                 processor_type: pentium III
                                                 hard_size: 10GB
                                                 main_memory: 64MB
                                                 monitor_size:17inch}}




Jae Kyu Lee                                                          PRICAI2000        17
                    XFDL (Extensible Forms Description Language)


<?xml version="1.0"?>
<XFDL version="4.0.1">
      <page sid="p1">
              ...
              <field sid="A"><label>Enter side length A:</label><value>3</value>
              </field>
              <field sid="B"><label>Enter side length B:</label><value>4</value>
              </field>
              <field sid="C">
                       <label>Hypotenuse length C is:</label>
                       <editstate>readonly</editstate>
                       <value content="compute">
                               <cval>5</cval>
                               <compute>
                               <![CDATA[
                                      A.value <= "0" || B.value <= "0"
                                      ? "" : sqrt(A.value*A.value + B.value*B.value)
                               ]]>
                               </compute>
                       </value>
              </field>
      </page>
</XFDL>




         Jae Kyu Lee                                                                   PRICAI2000   18
                 Example of RTL
        {{requisition-2000/6/28-1
           (requisition_section : ICEC )
           (item : student’s salary)
           (budgetary_source: type-P research fund)
           (account: P-1234)
           (amount: 1000)
           (permission:              )
           (RULE_TRIGGER: (<RTL>
                                       <RuleGroup>Budget Control</RuleGroup>
                                       <RuleTitle> budgetary_appropriateness</RuleTitle>
                                       <Condition>
                                         <trigger>budgetary_source</trigger>
                                                <access>new-value</access>
                                                <when>after</when>
                                       </Condition>
                                       <ReturnedResult>
                                         <var>permission</var>
                                       </ReturnedResult>
                                  </RTL)}}

              Issue of Synonym Dictionary



Jae Kyu Lee                                                                 PRICAI2000     19
      Advantages of XRML
    Aid Human’s Knowledge Processing:
       •   Conversion of XRML (including Rule Identification Language) to
           XML/HTML


    Aid Agent’s Knowledge Processing against Web page
       •   Triggering rule-based inferences in the agents, possibly from the
           Workflow Management System


    Maintain Consistency between Rule Base and Web page
       •   Aid the extraction(semi-automatically) of rules from XML
           maintaining consistency between them




Jae Kyu Lee                                                 PRICAI2000    20
           Roads Ahead
 for the Implementation of XRML


   Agreement on the Syntax of XRML 0.0

   Methodology of Knowledge Modeling

   Tool Implementation

   Illustration with Test Cases

   Implementation in the Web Ecology and
    Knowledge Management Systems.



Jae Kyu Lee                        PRICAI2000   21
              Roads Ahead

   From XML to XRML
   eCatalog and Directory Balancing Theory
   Configuration with Options
   Agents Based Commerce Architecture
   Ontology Initiatives in EC
   From Data Mining to Web Mining



Jae Kyu Lee                       PRICAI2000   22
    Taxonomy of Electronic Marketplaces
             and e-Catalogs


                                                            B2C       B2B

   Seller-centric Marketplace
     • Dell, Cisco, Amazon: Manufacturers, Retailers                 
   Intermediary-centric Marketplace
      • Vertical Net, Global Trading Web, ProceureNet,
                                                                     
         Manufacturing.Net, Industry.Net, Supermerce

   Buyer-centric Marketplace
     • GE, Shoppoint.
                                                                       




     Jae Kyu Lee                                         PRICAI2000   23
    Intermediary-Centric Marketplaces

                                                                       1472




                        1472




    Business
    Customer                                                         Business Supplier


                                 Intermediary’s
                 1472

                                 Electronic Mall    1472




  Consumers                                        Business Supplier

                                                                              SCM


              Customer’s                            Supplier’s
                                   Shared            Product
           Order Information
                               Product e-Catalog   Information


 The architecture of intermediary-centric marketplace is basically
 the same as that of seller-centric marketplace
 SCM aspect of B2B is necessary

   Jae Kyu Lee                                                   PRICAI2000     24
    AI Opportunity in e-Catalog

   Ontology Study on Thesaurus: Standard
    Terms, Synonym, Hierarchy
   One-to-One Marketing and
    Personalized e-Catalog
   Personalized Directory Balancing
    Theory and Implementation using Logic
    Programming




Jae Kyu Lee                        PRICAI2000   25
 Prototype of Balancing Directory




Jae Kyu Lee              PRICAI2000   26
              Balancing Theory of Directory
             Extract Subsets and Link to the Buyer’s Directory
              from the Global Directory

                                                      Buyer’s Level of Detail
...            Multiple       GD
               Supplier                                        MyCo
                                       Standard
                                       Directory
                                                      A           B         C

      B                   A




                                   C
 Expressed By Logic                                Occurrence of Unbalance State


          Jae Kyu Lee                                       PRICAI2000     27
Ingredients of Directory Balancing

      Measure the Degree of Balance
      Directory Definition Language
         - Representation in Logic
      Directory Manipulation Language
         - Implementation by Logic Programming
      Automatic Balancing Engine
      Consistency Maintenance with Global
       Directory


Jae Kyu Lee                                 PRICAI2000   28
  Directory Manipulation Language

 SPLIT UP “A” TO “MyCo”

                                                          Split_up(“A”, “MyCo”) :-
   LINK Child of “A” TO “MyCo”,       Logic Expression link_child_of(“A”, “MyCo”),
       DELETE “A”.                                           delete(“A”).


                   MyCo                                         MyCo
Split-Up

           A               b2    C              a3      a4    a5        b2       C

 a3        a4   a5        b3 b4 c1 c2        a1 a2      a1    a2       b3 b4 c1 c2

a1 a2      a1   a2                                    a6 a7
        a6 a7                                           a8 a9
           a8 a9


  Jae Kyu Lee                                                      PRICAI2000        29
Directory Manipulation Language
 MERGE “b2” AND “C” UNDER “New” UNDER “MyCo”
    LINK “New” TO “MyCo”,
     LINK “b3” TO “New”,                   Merge ([“b2”, “C”], “New”, “MyCo”) :-
                                   Logic
     LINK “b4” TO “New”,                     link_to(“New”, “MyCo”),
     LINK “c1” TO “New”,                     link_to(MergedlList, “New”).
     LINK “c2” TO “New”,
     DELETE “b2”,
     DELETE “C”.

               MyCo                                           MyCo

                           Merge
  a3      a4          b2      C                  a3      a4            New

a1 a2     a1      b3 b4 c1 c2                 a1 a2      a1      b3 b4 c1 c2
        a6 a7                                         a6 a7
          a8 a9                                          a8 a9

 Jae Kyu Lee                                                         PRICAI2000      30
       Search toward more Balanced Directory


                                                                1.20
                                                                1.00
   Balancing of Template5-12
                                                                0.80
                                                                0.60
                                                                0.40
                                                                0.20
                                                                0.00
                                                                        Template5-12   Template5-7    Template5-2   Template5-1
0.80                                                                       1.11           1.04           0.71          0.00
                                                                 T-UD
                                                                 D-UD      0.74           0.60           0.33          0.00
0.60                                                             W-UD      0.38           0.44           0.38          0.00

0.40

0.20

0.00
        Template5-8   Template5-3   Template5-2   Template5-1
 T-UD      0.40          0.29          0.71          0.00
                                                                         Balancing of Template5-8
 D-UD      0.23          0.13          0.33          0.00
 W-UD      0.17          0.17          0.38          0.00



        Jae Kyu Lee                                                                                  PRICAI2000           31
              Roads Ahead

   From XML to XRML
   eCatalog and Directory Balancing Theory
   Configuration with Options
   Agents Based Commerce Architecture
   Ontology Initiatives in EC
   From Data Mining to Web Mining



Jae Kyu Lee                       PRICAI2000   32
           Comparison e-Broker and Software Agent

            Online Site                                                 : Internet
            Search
                                                                        : EDI, Fax, Phone
 mfr                           e-Broker                  Customers
                                                                        : Visit
                                                                        : Warehouse


 Typical e-Brokers Service in the U.S.A.
Compmay                URL
                                                         Korea
CompareNet             www.compare.net
Excite’s Jango.com     www.jango.com
Cadabra.com            www.cadabra.com                    Www.b2bplace.co.kr
Mysimon                www.mysimon.com                    Shopbinder.com
Altavista              Shopping.altavista.com/compare
Personalogic           www.personalogic.aol.com           Shopmadang.com
BottomDollar           www.bottomdollar.com               Webnara.co.kr
Netbuyer               www.zdnet.com/computershopper      Yavis.com
ComputerEsp(Uvision)   shopper.cnet.com
PriceScan              www.pricescan.com                  Am.co.kr
KillerApp              www.killerapp.com

         Jae Kyu Lee                                                 PRICAI2000       33
  Current Status Comparison Shopping



     Simple Configuration
     Graphic Displays
     Preference Based Scoring
     Tabular Comparison




Jae Kyu Lee                      PRICAI2000   34
Jae Kyu Lee   PRICAI2000   35
Jae Kyu Lee   PRICAI2000   36
Jae Kyu Lee   PRICAI2000   37
Jae Kyu Lee   PRICAI2000   38
Jae Kyu Lee   PRICAI2000   39
Jae Kyu Lee   PRICAI2000   40
Opportunities in Comparison Shopping
Current Status
   Configuration
   Graphic Displays
   Preference Based Listing
   Tabular Comparison


Opportunities
   Configuration with Options:
    - CBR: Find the Best Standard Model
    - CSP: Search toward the best options

    Multi-vendor Opportunity and Multi-Agent
    based Configuration and Order Processing
   Buyer Agents

Jae Kyu Lee                                 PRICAI2000   41
               Buyer Agents as Mediators in EC

      Online Shopping framework with representative examples of agent mediation


                         Persona      Firefly    Bargain      Jango      Kasbah      Auction   T&T   UNIK
                          Logic                  Finder                                Bot
Need identification

Product brokering                                                                             
Merchant brokering
                                                                                                  
Negotiation                                                                                  
Payment and delivery

Service and evaluation



                      (Maes, March 1999/, Vol. 42. No. 3 Communication of the ACM)



        Jae Kyu Lee                                                                   PRICAI2000      42
              Roads Ahead

   From XML to XRML
   eCatalog and Directory Balancing Theory
   Configuration with Options
   Agents Based Commerce Architecture
   Ontology Initiatives in EC
   From Data Mining to Web Mining



Jae Kyu Lee                       PRICAI2000   43
            Buyer-Centric Electronic
                 Marketplaces
                                                    Bid               Business
                                                                      Supplier


                                                                       Business
                                Buyer’s                                Supplier
                             Electroic Mall
 Business
  Buyers

                       Buyer’s
                  Requesting Products           Supplier’s
                       Catalog                Bid Information




Reverse Auction, and eProcurement Systems



    Jae Kyu Lee                                                 PRICAI2000   44
         Electronic Reverse Auction

                                             GE TPN(Trading Process Network)
                                                Reengineer purchase process
   Mfr                                          Purchase on the Internet
                                                 $10B(‘97.10) $50B(2000)
                                                60% of crews relocated , 30%
                   Requests for   Business       reduced
Distributor
                      Bids         Buyer        5~20% reduced price
                                                Bidding cycle time reduced
                                                New vendors
   VAR
                                                Open TPN to other buyers




 Purchase Reengineering
 Cultivate New Market

     Jae Kyu Lee                                                 PRICAI2000     45
        Status of Biddings Sites and
         Necessity of Seller Agents
Nations    eMarketplace                    URL                 Controller     Fee     Maturity
                                                                                                     : Operational
               CBDNet           http://cbdnet.access.gpo.gov   government     Free

                 GPO            http://www.access.gm.gov       government     Free
                                                                                                     : Information
                 COS                 http://cos.gdb.org        government    Charge

                                                               Third Party
               BIDCAST            htpp://www.bidcast.com                     Charge
                                                               Corporation                           : Under
 U.S.A         BIDLINE             http://www.bidline.com           “        Charge                  Construction
               BIDNET             http://www.bidnet.com             “        Charge

          Federal Marketplace    http://www.fedmarket.com           “        Charge              Over 2,100 bid
               GOVCON             http://www.govcon.com             “        Charge
                                                                                                 opportunity URLs
                                                                                                 in Federal
                 LDC              http://www.id.com                 “        Charge
                                                                                                 Marketplace
                MERX            http://www.merx.cebra.com           “        Charge              (Source:Bidengine.com)
Canada                                                                                           State & local :1150
                 BIDS           http://www.bidservices.com          “        Charge
                                                                                                 Federal : 550
  EU             EPIN              http://epin1.epin.ie        Consortium    Charge
                                                                                                 International : 400
 Hong
 Kong
                 GSD            http://www.info.gov.hk.gsd     government     Free




  Jae Kyu Lee                                                                                PRICAI2000           46
              Types of Agents
    Multi-agent Negotiation System: Research Stage
      • Tete-a-Tete, Kasbah
      • Contract Protocol Definition is necessary
    Mobile Agent: Data Collector with Simple Parser
    Assistant Agent:
     1) Voice Recognition & Synthesis: Secretary
         Portico (www.genmagic.com/portico)
     2) Natural Language Processing: Template Answers
        (www.extempo.com, www.neuromedia.com)
    User Interface Agent
     AdEater(www.cs.ucd.ie/staff/nick/research/ae)
    Intelligent Agent
      • Personalogic
    Filtering: Maxims(e-mail), Firefly (Collaborative Filtering)



Jae Kyu Lee                                         PRICAI2000   47
                  Agent-based EC:
             Contract Type and Protocol
         Requirement of Human Customer          Requirement of Human Customer



                           Buyer C1               Buyer C0




        Seller R1
                                                                  Seller R4
                           Seller R2             Seller R3




Final Approver1
                  Final Approver2      Final Approver3          Final Approver4

        Jae Kyu Lee                                             PRICAI2000    48
      Architecture of UNIK-AGENT
                             AGENT


Problem Solver                     Communication Controller
                                          Outgoing Msg
Problem Solving Manager
                                  Incoming Msg
                                                                  Message Base
                                            Message Manager                      Directory



                                           Directory Consulting
    Solution Engines                                                             Other
                                                                                 Agents
                                          Individual Messages


Data Base        Knowledge Base           Message Queue Mgt.


                                                 Message Gate



 Jae Kyu Lee                                                      PRICAI2000      49
 Three Layers of EC Messages

   Agent Communication Language Layer
     • Domain independent communication language
       among agents (e.g. Knowledge Query and
       Manipulation Language: KQML)


   Electronic Commerce Layer
     • Message types and items for agent based commerce


   Product Specification Layer
     • Representation the specification of products



Jae Kyu Lee                                   PRICAI2000   50
                                                    Message Standard in Three Layers
                                                                                                 (evaluate
                                                                                                             :sender                         C1
                                                                                                             :receiver                       R1 R2 R3 R4 R5
                                                                                                             :reply-with                     msg_960924_1
                                                                                                             :ontology                       Agent Based Commerce
Agent Ccommunication Language Layer
                                                                                                             :language                       UNIK-OBJECT
                                                                                                             :content
                                                                                                                           ((title             RFP)
                                                                                                                           (contract_ID        contract_960924)
                                                                                                                           (contract_type
                                                                                                                                             (number_of_bid_round              1)
                                                                                                                                             (competitor_price_referable       not_referable)
                                                                                                                                             (announced_estimated_price_limit enforced)
                                                                                                                                             (number_of_proposal_for_each_bidder 1)
                                                                                                                                             (bid_price_change                  not_allowed)
                                                                                                                                             (bid_price_open_time               at_predetermined_time)
                                      Electronic Commerce Layer




                                                                                                                                             (buyer_bidder_prenomination        prenominated_by_customer)
                                                                                                                                             (human_involvement                 enforced)
                                                                                                                                             (bidding_price_type                total_amount)
                                                                                                                                             (buyer_rule_for_selection_of_successful_bidder min_price)
                                                                                                                                             (buyer_nego_between_price_&_spec not_allowed))

                                                                                                                           (bid_time
                                                                                                                                            (end_time 96/10/01/11/00))


                                                                                                                           (requirement
                                                                  Product Specifications layer




                                                                                                                                             (payment_method                    credit_card)
                                                                                                                                             (delivery_method                   postal)
                                                                                                                                             (delivery_date                     Oct 14, 1996)
                                                                                                                                             (products
                                                                                                                                                              (item_name                       PC)
                                                                                                                                                              (quantity                        1)
                                                                                                                                                              (amount                          (<= 2500))
                                                                                                                                                              (specifications
                                                                                                                                                                 (main_memory                    (>= 16MB))
                                                                                                                                                                (hard_disk                       (>= 840MB))
                                                                                                                                                                (processor                       (>= Pentium90)))))))




Jae Kyu Lee                                                                                                                                                                     PRICAI2000                    51
              Roads Ahead

   From XML to XRML
   eCatalog and Directory Balancing Theory
   Configuration with Options
   Agents Based Commerce Architecture
   Ontology Initiatives in EC
   From Data Mining to Web Mining



Jae Kyu Lee                       PRICAI2000   52
    Ontology Initiatives in EC
   ACL in XML Form

   Biz Talk (Microsoft)

   OBI (Open Business Interface)

   CBL (Common Business Language)

   ebXML

Goal of EC: Make the complex procedure simple and
    standardized

Attitude of AI: Keep the problem as complex as it was.
    We will show that only AI can solve this problem.

         “You seem smart, but I will not buy it.”

Jae Kyu Lee                                     PRICAI2000   53
OBI (Open Buying on the Internet)

                                                                               Supplier
                                                                             Catalog Mgmt.

  Requisitioner
                                           a                                Price Information
                                                                                   Mgmt.

                                                                               Order Entry &

                                           
                                                                             Inventory Mgmt.
   b 
                                                                            
      Buying
    Organization
       Supplier Search
                                                                             Payment
     Requisitioner Profile
           Mgmt.
                                                                            Authority
         Approval.
                             1. Connect to BO's Web Server and Select a Hyperlink to SO's
                             catalog.
                             2. Authenticate Requisitioner using Digital Certificate
                             3. OBI Order Request
                             4. Add Administrative Information
                             5. OBI Order
                             6. Obtain Credit Authorization
                             7. Issue Invoice and Receive Payment

Jae Kyu Lee                                                                      PRICAI2000      54
                          ebXML
       Vision of EDIFACT
         • A single set of internationally agreed upon
           technical specifications that consist of
           common XML semantics and related document
           structures to facilitate global trade

       Project Teams
         •    Requirements
         •    Business Process
         •    Technical Architecture
         •    Core Components
         •    Transport/Routing and Packaging
         •    Registry and Repository
         •    Technical Coordination and Support
         •    Marketing, Awareness and Education




Jae Kyu Lee                                        PRICAI2000   55
              Roads Ahead

   From XML to XRML
   eCatalog and Directory Balancing Theory
   Configuration with Options
   Agents Based Commerce Architecture
   Ontology Initiatives in EC
   From Data Mining to Web Mining
    - CRM and SCM

Jae Kyu Lee                       PRICAI2000   56
       Data Mining Taxonomy
                                               Source: http://www.crm-forum.com
                           Data Mining

      Verification Driven DM         Discovery Driven DM

SQL      SQL Generator         Description                 Prediction

Query Tools                                              Statistical
                                      Classification
                  Visualization                          Regression
       OLAP
                    Clustering                  Decision Tree
                   Association
                                                Rule Induction
        Sequential Association
                                                Neural Network
                    Distillation

 Jae Kyu Lee                                           PRICAI2000          57
   Taxonomy of Web Mining

                    Web Mining

 Web Content       Web Structure         Web Usage
   Mining             Mining              Mining

• Content          • Link Analysis   •   Personalization
  Classification   • Information     •   Cross Sell/Up Sell
• Information        Search Aid      •   Segmentation
  Filtering                          •   Dynamic Ads Mgt.
• Recommendation                     •   And more and
                                         more



 Jae Kyu Lee                                  PRICAI2000      58
               Web Mining
             Everywhere in EC
                                 Customized Catalogue
                            Browse
                                    Customized Banner                          Pay

      Navigation Prediction                                  Fraud Detection
         Entry                 Home                 Login        Add to Cart         Select
   Sequential Event Mining
                                                             Cross Sell/Up Sell
                                                  Register
                                             Market Basket Analysis
                                    Content Based Filtering
                              Search

   Source:
                                    Collaborative Filtering
   Menace, D. and A. F. Almeida, Scaling for E-Business



Jae Kyu Lee                                                               PRICAI2000     59
SCM(Supply Chain Management)


 Supplier         Mfrer    Logistic   Distributor    Customer




              • Reduce Cycle Time
              • Reduce Inventory
              • Reduce Costs in Value Chain



Jae Kyu Lee                                         PRICAI2000   60
          Bullwhip Effects in SCM


     Increased Fluctuation in the Upstream

                     Pampers Supply Chain




3M             P&G     Wholesales   Retailers   Customers     Babies



         Upstream                                Downstream




 Jae Kyu Lee                                          PRICAI2000   61
     Point-to-Point and Hub


      Point to Point   E-Hub Collaboration




Jae Kyu Lee                     PRICAI2000   62
              Wal-Mart’s RetailLink:
                Data Mining Aid
   Wal-Mart: World’s largest discount retailer
    (sales: $224 bil/year as of 1999)
   2,400 stores and 450 Sam’s Club in the U.S.A.
    720 units internationally; 7,000 suppliers
   Traditionally, retailers and suppliers forecasted separately
     Excessive inventory, out-of-stock, lost opportunity to
    suppliers
   RetailLink Data Warehouse for 2 years of transaction data (101
    terabytes)
   120,000 data mining questions per week
   Retail industry’s savings are projected to be $150-$250 billion
    per year

Jae Kyu Lee                                         PRICAI2000     63
             Collaborative Forecasting and
            Replenishment with DW and DM

       Wal-Mart
                          Operational
                            System                                 Warner-Lambert

                                                         EDI

                       Data warehouse                                   ERP


                                           Internet
                                                                   Manufacturing
                         RetailLink                                    Plan

                        Sales data about
                                                       WWW
                         W-L Products


                                                      Review and
Inventory                 Forecast                    Comments
   Plan
                                                                       Planner
             Planner

      Jae Kyu Lee                                              PRICAI2000        64
                 Conclusion
   AI has rich opportunities in EC
     •   XRML
     •   Personalized Directory Balancing
     •   Configuration with Options
     •   Architecture of Agents in EC
     •   Ontology Initiatives ???
     •   Web Mining in CRM and SCM

   At least, a group of people in AI society needs to
    commit to EC, not to demonstrate the possibility
    of AI but to really solve the problem.
   AI community needs to pay more attention to the
    comparative validation.




Jae Kyu Lee                                 PRICAI2000   65

				
DOCUMENT INFO