A knowledge-based system that gives better price recommendations than a good manager : a dissertation submitted to the Faculty of Business Studies, Massey University in fulfilment of the requirements for the degree of Doctor of Philosophy

Copyright is owned by the Author of the thesis. Permission is given for

a copy to be downloaded by an individual for the purpose of research and

private study only. The thesis may not be reproduced elsewhere without

the permission of the Author.

A KNOWLEDGE-BASED SYSTE M

THAT GIVES BETTER PRICE RECOMMENDATI ONS THAN A GOOD HANAGER

Ching Biu Tse August 1991

A dissertation submitted to the Faculty Massey University, in,fulfilment of the·

degree of Doctor of Philosophy.

of Business requirements

Approved by:

studies, for the

TO WHOM IT _MAY CONCERN

Thi s. i s to state that the research carried out f or my Ph . D . the s i s entitled " A Knowledge-based System that G ives B etter Price Recommendations than a Good Manager" in the Marketing Department at Massey Un ivers i ty , New Zealand , i s a l l my own work .

Thi s i s a lso t o certi fy that the thesis mater i a l has not been used.for any other degree .

Ching B iu Tse

Dr Anthony C Lewis: Supervisor

Date :

ACKNOWLEDGEMENT

I would l ik e to express my deep feel ings o f grati tude to the supervi s or of my study , Dr Tony Lewis , and a lso to my adv isor , Mr Ray Kemp , who have both been very generous in providing me with their valuable guidance during this research endeavour . The ins ight ga ined from i nteraction with them w ill conti nue to be a va luabl e asset throughout my professional life . ^I would a l s o

l ike t o thank M r Don Ess lesmont , Professor Phi l Gendal l , D r Mike Brennan� Mr Robert Langton , Ms Pascale Quester and Dr Peter Andrea . for their helpful suggestions . ' ^I am indebted to a l l the manager s participating in the knowledge engineering part of thi s study f or providing me w ith valuable rules

h

f how pr ice deci sions should be made and the values of the membership functions used in the system . I am a l s o grateful to Ms M . M . Hilder o f the Modern '

Languages Department and Ms L . stock for their help in touching up the Eng l i sh in thi s thes i s .

ABSTRACT

The obj ective of this research was to investigate whether i t i s pos s ib l e t o con struct a computer system that provides good pri ce recommendation s for marketers in the export wool industry . The computer system is a program that stores and makes u s e o f . expert knowledge in the course of providing recommendation s to a user .

" Good'l price recommendations are ones that c losely resemble those '

of a consensus of experts .

Fuz z y logic, a method o f inexact reason ing, is used to derive recommendation s from inexact rules in the system . The system a l lows users to input vague express ion s used in natural language , tran slates them into n on fuzzy va lues and then performs a set of I

operations on them to produce a nonfuz zy recommended price . _'

The e f f ectiveness of the system in making good price 'dec i s ion s was tested by comparing the system ' s recommendations w ith consen sus recommendat i ons from a panel of experts . The correlation coefficient between the system's recommendations and the consensus recommendations was found to be higher than .any o f the correlation coe f f icients between the individua l manager's f ir st recommendation s and the consensus recommendations . Thi s suggests that the system constructed i s capable o f providing good

I

price recommendations .

i i

Acknowl edgement Abstract

List of Tables List of Figures Chapter 1

1 . 1 1 . 2 Chapter 2

2 . 1 2 .1.1

2 . 1 . 1 . 1 2 . 1 . 1 . 2 2 . 1 . 1 . 3 2 . 1 . 2 2 . 1 . 2 . 1 2 . 1 . 2 . 2 2 . 1 . 3 2 . 1 . 4 2 . 2 2 . 3 2 . 3 . 1 2 . 3 . 1 . 1 2 . 3 . 1 . 2 2 . 3 . 1 . 3 2 . 3 . 1 . 4 2 . 3 . 1 . 5 2 . 3 . 1 . 5 . 1 2 . 3 . 1 . 5 . 2 2 . 3 . 2 Chapter 3

3 . 1 3 . 1 . 1 3 . 1 . 2 3 . 2

TABLE OF CONTENTS

Page i i i v i v i i Introduction

Proj ect Description . . . ^{1 • • • • • • • • • • • • • • • • •} Organ i z ation of the Thes i� . . . ^. Kno� ledge-based Systems in Marketing

Structure of K nowledge-based Systems

The Knowledge Base . . . . Dec i s i on Rules . . . . Semantic Networks . . . ^. Frames . ^. . ^{. . . . .} . ^{. .} . . . ^. . . . ^. . ^{. . . . . . . . . . .· . . . .} The I nference Engine ^.

.1.

^{• •: • • • • • • • • • • • • •· • • • •}

Backward Cha ining . . . . Forward Chaining . . . ^{. .} . . . · . . . . The U�er Interface . . . ^, . . . ^. . ^{. . . . .} . ^{. . .} . ^. . Conclusion . . ^{. . . .} . . . ^. . . . ^. . . . ^{. . . . . . .} . ^{. . .}· . . ^. Review of Knowledge-based Systems in

Market ing . . . ^. Fuz zy Logic . . . ^{; . . . .} . ^. . . ^. . ^. . ^{. . . . . . . . . .} The Fuzzy Logic Approach . . . ^. The Conj unction Operator . . . ^. The Disjunction Operator . . . . The Negation Operator ^{. . . . . . . . . . . • . . . • . . . .} The Condition a l Operator ^{. . . . • . . .}: ^{. . • . . . . .}

The Composition Operator . . . ^. Hedg ing . . . ^. Appl ication of the Compos ition

Operation . . . ^. Conc lus ion . . . ^{; . .} . . . . Knowledge Engineering and Method

Knowledge Eng ineer ing ^{. . . . . . . . . . . • . .} : . ^. . ^{. . .} I ntroduction ^{. . .} . ^{. . . . . . . .} .^. . ^{. . . . .} . ^. . ^{· . . . . . .} Choice of Knowledge Engineering Methods i n the Current Study ^{. . . .} . ^{. . . . . ... . . .1 • • • • • •} The Delphi Technique and its Appl ica� ion in K nowledge Engineering . ^. . ^{. . .} . ^{. . . .}� .... ^.

1 4

7 6 7 1 0 1 1 1 2 1 2 1 3 1 4 1 4 1 5 2 3 2 3 2 7 2 8 3 0 3 1 3 3 3 4 3 6 3 7 4 0 4 0 4 5 4 6

3 . 3 3 . 3 . 1 3 . 3 . 1 . 1 3 . 3 . 1 . 2 3 . 3 . 1 . 2 . 1 3 . 3 . 1 . 2 . 2 3 . 3 . 1 . 2 . 3 3 . 3 . 2 3 . 3 . 2 . 1 3 . 3 . 2 . 2 3 . 3 . 2 . 2 . 1 3 . 3 . 2 . 2 . 2 3 . 3 . 2 . 2 . 3 3 . 3 . 3 3 . 3 . 3 . 1 3 . 3 . 3 . 2 3 . 3 . 3 . 2 . 1 3 . 3 . 3 . 2 . 2 3 . 3 . 3 ^• 2 . 3 3 . 1

Chapter 4 4 . 1 4 . 2 4 . 2 . 1 4 . 2 . 2 4 . 2 . 3 4 . 2 . 4 4 . 2 . 5 4 . 3 4 . 3 . 1 4 . 3 . 2 4 . 4 4 . 4 . 1 4 . 4 . 2 4 . 4 . 3 4 . 4 . 4

Chapter 5 5 . 1 5 . 2 5 . 2 . 1 5 . 2 . 2

The Appl ication o f the Delphi Technique in the current Study ^{. . • . . . . . . . . . . . . • . . . . .} The P ilot study ^{. .} . ^{. . .} . ^. . . ^. . ^· . . ^. . ^{. . . . . . .} � . . ^.

Method Used in the P i lot Study . . . . Results o f the Pilot Study . . . . Process of Price Deci s ion Making . . . . Important Fuz zy Factors in the Dec i s ion Making Process ^{. . . . . . . . . . . . . . . . . . . . . . . . . . .·} Dec i s ion Rules Used in the Decision

Making Process . . . . The Fieldwork for Initial Responses . . . . Method . . . . Results o f the Fieldwork . . . . Manager s ' Recommendations G iven the '

Scenarios ^. . . . ^. . . . ^. . . ^·i ^{• • • • • . • • • • • • • • • • •}

Initia l Membership Functions . . . . Selection of Managers for Delphi Sessions . The De lphi Sessions . . . . Method ^·. � . . . . Results o f the De lphi Study . . . ^. Decision Ru les . . . ^. Membership Functions . . . ^. Consensus Deci s ions . . . ^; . . . ^. System Evaluation . . . .

�.: ...

^{; . . . .}

System Specification

Phi losophy for the Construct ion of TZ . . . . Knowl'edge Stored in TZ' . . . · . . . ^. Membership Functions o f Primary Fuz zy · Variables . . . . Relational Matrices . . . . Formulae for Hedging Functions . . . . Relative ly Constant Facts . . . ^. Accounting Data . . . ^. structure and Inference Mechanism in TZ . . Components of TZ ^{. . . . . . . . . . . . . . . . . • . . . . . . .} The I nference Mechanism . . . • . . . .

Prolog : The Programming Language Used in T Z ^{• • • . • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •} I ntroduction to .Prolog ^{. . . . . . . . . . • . . • . . • . •} Knowl edge Representation in Pro log .: ^. . ^. . . ^. Inference Procedure in Prolog . . . : ^{. . . � . .}

Reasons for Us ing Prolog as the ¹

Programming Language . . . : ^. . ^. . ^{. .} TZ , the Knowledge-based System for �rice

Dec i s ion Making ^I

5 0 5 0 5 0 5 2 5 2 5 2 5 4 5 4 5 4 6 1 6 1 6 1 6 1 6 2 6 2 6 4 6 4 64 66 6 9 7 2 7 5 7 5 7 5 7 6 7 7 7 8 8 0 8 0 8 2 8 9 8 9 9 0 9 2 9 6

Samp l e Run of T Z ^{· · · · · · · · · · · · · · · · · · �}I ^{· · · · · ·} 1 0 1 Eva luat ion of T Z ^{. . . . . . . . . . . . . . . . • .} ; . . . 1 1 0 The Objective Revis ited ^{. . . . . . . . . . • . . . . . . .} 1 1 0 System Evalua:tion . . . . . . . • . . . . . . . . 1 1 1

' .

iv

Chapter 6

Refer(:mces Append ix 1 Appendix 2 Appendix 3 Appendix 4

Appendix 5

Conclusions and Directions for Future ·

' Research . . . 1 1 3

\

1 1 5 1 1 9 1 2 3 1 3 9 1 5 6 1 6 0

Table

3 . 1 3 . 2 3 . 3

L I ST OF TABLES

D escription

Importance of Fuz zy Factors

Response Rates of the Delphi Sess ions I

Consensus Recommendations to Problem Scenarios

Page

5 3 6 3, 67

Figure·

2 . 1 2 . 2

2 . 3 3 . 1 4 . 1 4 . 2 4 . 3

4 . 4 4 . 5

L I S T OF FIGURES

Descr ipt i'on

Knowledge-h ased System Components

Tree- l ike Representation of causes o f Fal l i n Market Share

Relat iona l Matrix

Stages of Knowledge Acquisition

Re lational Matrix for Ru le 1

Structure and Inference Mechanism o f TZ Fl ow Diagram Showing the Inference

Mechanism in T Z

Tree Representat ion of the Above Example Tree Solution of "member ( 2 , [ 1 , 2 , 3 ] ) ? " I,

'

Page

7 1 6

33·

42

7 6 8 0 8 2 9 6 9 8