1/21

Chapter 2
Overview of the Operations Research
Modeling Approach

Heesang Lee
SKKU
 2/21

The Modeling Process

1. Define the problem of interest and gather relevant

data.
2. Formulate a mathematical model to represent the
problem.
3. Develop a computer-based procedure for deriving
solutions to the problem from the model.
4. Test the model and refine it as needed.
5. Prepare for the ongoing application of the model as
prescribed by management.
6. Implement.

HSLee@SKKU
 3/21

2.1 Define the Problem & Gathering Data

 Consider the frame of the problem you are attack-

ing – should you be thinking bigger? Should you
focus at a more detailed level?
 Objectives: What are you trying to achieve? How
will you evaluate your alternatives? Do you have any
constraints?
 Alternatives: What you are choosing between.
Something you can control. Try to generate alterna-
tives that best match objectives.
 Information: Data and uncertainties. Facts. Things
that you can’t control (or you don’t want to con-
sider within this problem).
HSLee@SKKU
 4/21

OR Teamwork
 OR team normally works
 in an advisory capacity
 set appropriate objectives
 for entire organization
 for long-run profit (= revenue – cost) maximization
 surprisingly large amount of time gathering relevant
data about the problem

HSLee@SKKU
 5/21

Examples
 Example Case - Merrill Lynch (Textbook pp.11~12)
 to a complete overhaul in how it charged for its services, rang-
ing from a full-service asset-based option to a low-cost option
for clients wishing to invest online directly
 data collection and processing  200 gigabyte client database
involving 5 million clients, 10 million accounts, 100 million
trade records, and 250 million ledger records
 required merging, reconciling, filtering, and cleaning data from
numerous databases
 one-year increase of nearly $50 billion in client assets held and
nearly $80 million more revenue.

HSLee@SKKU
 6/21

2. Modeling
 문제를 분석에 적합한 형태로 모형 정립
 전통적인 OR 접근 방법은 문제의 본질을 표현하는
수학적 모형을 만듬
 모형은 연구를 위한 주제의 핵심을 추상화하거나 ,
관계를 보여주거나 , 분석을 촉진하게 하는데에 필수적
 모형 : 이상적 현실의 표현
 일상의 필수적인 부분
 모형의 예 :
 모형 비행기 , 초상화 , 지구본
 원자 구조 모형 , 유전자 구조 모형
 운동의 물리적 법칙이나 화학적 반응을 설명하는 수학적 방정식
( 예 : F = ma, E = mc2)
 그래프 , 조직도 , 기업의 회계 장부

HSLee@SKKU
 7/21

Mathematical Modeling
 Decision variable: 결정변수 (decision variable) x1, x2, … , xn
 적절한 성능 척도 ( 예 : 기업의 수익 ) 를 이들 결정변수의 함수
형태로 표현한다 ( 예 : P = 3x1 + 2x2 + . . . + 5xn).
 결정변수에 주어질 제약조건은 수학적으로 , 보통은 항등식이나
부등식으로 , 표현이 가능 ( 예 : x1 + 3x1x2 + 2x2 ≦ 10).
 The coefficients of the constraints are often called the technological
coefficients.
 The number on the right-hand side of the constraint is called the
constraint’s right-hand side (or rhs).
 Sign Restrictions: nonnegative decision variables (xi ≥ 0), or urs
(unrestricted in sign) decision variables.

HSLee@SKKU
 8/21

Modeling

Concepts of linear function and linear inequality:

 Linear Function: A function f(x1, x2, …, xn) of x1, x2, …, xn is a

linear function if and only if for some set of constants, c1, c2,

…, cn, f(x1, x2, …, xn) = c1x1 + c2x2 + … + cnxn.

 For example, f(x1,x2) = 2x1 + x2 is a linear function of x1 and x2, but f(x1,x2)
2
= (x1)  x2 is not a linear function of x1 and x2.

 For any linear function f(x1, x2, …, xn) and any number b, the

inequalities f(x1, x2, …, xn) £ b and f(x1, x2, …, xn) ³ b are linear
inequalities.

HSLee@SKKU
 9/21

Modeling

The feasible region of an LP is the set of all points satisfying all

the LP’s constraints and sign restrictions.

Example: LP Constraints
 x1 = 40 and x2 = 20 are in the feasible
2 x1 + x2 ≤ 100 (factory 1 constraint)
region since they satisfy all the
x1 + x2 ≤ 80 (factory 2 constraint) constraints.
x1 ≤ 40 (product 1 demand constraint) On the other hand, x1 = 15, x2 = 70 is

x1 ≥0 (sign restriction) not in the feasible region because this
x2 ≥ 0 (sign restriction) point does not satisfy the carpentry
constraint [15 + 70 > 80].

HSLee@SKKU
 10/21

Modeling
 An optimal solution is a feasible solution that has the
optimal objective function value.
 For a maximization problem, an optimal solution to an
LP is a point in the feasible region with the largest ob-
jective function value.
 Similarly, for a minimization problem, an optimal solu-
tion is a point in the feasible region with the smallest
objective function value.

HSLee@SKKU
 11/21

Modeling Examples
 Example Case - Continental Airlines (Textbook p.14)
 One of the most challenging problems faced by airlines is how
to quickly reschedule the assignments of crews to flights when
flight delays or cancellations occur because of inclement
weather, aircraft mechanical problems, or crew unavailability.
 Because the airline has many thousands of crews and daily
flights, the model needed to be huge to consider all possible
pairings of crews with flights. Therefore, the model has millions
of decision variables and many thousands of constraints.
 In its first year of use (mainly in 2001), the model was applied
four times to recover from major schedule disruptions (two
snowstorms, a flood, and the September 11 terrorist attacks).
 This led to savings of approximately $40 million. Subsequent
applications extended to many daily minor disruptions as well.

HSLee@SKKU
 12/21

Modeling Examples
 Example Case - Rijkswaterstaat (Textbook pp.14~15)
 The Netherlands government agency responsible for water control
and public works
 Development of a new national water management policy
 The new policy saved hundreds of millions of dollars in investment
expenditures and reduced agricultural damage by about $15 million
per year, while decreasing thermal and algae pollution.
 Rather than formulating one mathematical model, this OR study de-
veloped a comprehensive, integrated system of 50 models
 Furthermore, for some of the models, both simple and complex ver-
sions were developed.
 The simple version was used to gain basic insights, including trade-
off analyses.
 The complex version then was used in the final rounds of the analy-
sis or whenever greater accuracy or more detailed outputs were de-
sired. The overall OR study directly involved over 125 person-years
of effort (more than one-third in data gathering), created several
dozen computer programs, and structured an enormous amount of
data.

HSLee@SKKU
 13/21

2.3 Deriving Solutions from the Model

 Algorithms: systematic solution procedures

 Heuristic procedures: intuitively designed procedures
that do not guarantee an optimal solution to find a
good suboptimal solution.
 In recent years, great progress has been made in de-
veloping efficient and effective metaheuristics (Ch.13 in
this class) that provide both a general structure and
strategy guidelines for designing a specific heuristic
procedure to fit a particular kind of problem. (see ch.
13)

HSLee@SKKU
 14/21

Solutions Examples

 Example Case - Rijkswaterstaat (Textbook p.17)

 did not conclude by recommending just a single solution. Instead, a
number of attractive alternatives were identified, analyzed, and com-
pared.
 The final choice was left to the Dutch political process, culminating
with approval by Parliament.
 Using sensitivity analysis (a post-optimality analysis) certain parame-
ters of the models represented environmental standards.
 Sensitivity analysis included assessing the impact on water manage-
ment problems if the values of these parameters were changed from
the current environmental standards to other reasonable values.
 Sensitivity analysis also was used to assess the impact of changing the
assumptions of the models, e.g., the assumption on the effect of fu-
ture international treaties on the amount of pollution entering the
Netherlands.
 A variety of scenarios (e.g., an extremely dry year and an extremely
wet year) also were analyzed, with appropriate probabilities assigned.

HSLee@SKKU
 15/21

2.4 TESTING THE MODEL

 model validation is testing and improving a
model to increase its validity
 retrospective test
 Test involves using historical data to reconstruct the
past and then determining how well the model and
the resulting solution would have performed if they
had been used.
 Comparing the effectiveness of this hypothetical per-
formance with what actually happened then indi-
cates whether using this model tends to yield a sig-
nificant improvement over current practice

HSLee@SKKU
 16/21

Testing Example

 Example Case - IBM (Textbook p.19)

 integrate its national network of spare-parts inventories to im-
prove service support for IBM’s customers.
 This study resulted in a new inventory system that improved
customer service while reducing the value of IBM’s inventories
by over $250 million and saving an additional $20 million per
year through improved operational efficiency.
 A particularly interesting aspect of the model validation phase
of this study was the way that future users of the inventory
system were incorporated into the testing process.
 Because these future users were skeptical about the system be-
ing developed, representatives were appointed to a user team
to serve as advisers to the OR team.
 After a preliminary version of the new system had been devel-
oped (based on a multi-echelon inventory model), a pre-im-
plementation test of the system was conducted.
 Extensive feedback from the user team led to major improve-
ments in the proposed system.

HSLee@SKKU
 17/21

2.5 PREPARING TO APPLY THE MODEL

 If the model is to be used repeatedly, the next step is to install a
well documented system for applying the model as prescribed by
management.
 This system will include the model, solution procedure (including
postoptimality analysis), and operating procedures for implementa-
tion.
 Then, even as personnel changes, the system can be called on at
regular intervals to provide a specific numerical solution.
 This system usually is computer-based. In fact, a considerable
number of computer programs often need to be used and inte-
grated.
 Databases and management information systems may provide up-
to-date input for the model each time it is used, in which case in-
terface programs are needed.
 After a solution algorithm is applied to the model, additional
computer programs may trigger the implementation of the results
automatically.

HSLee@SKKU
 18/21

PREPARING Example

 Example Case - Continental Airlines (Textbook p.20)

 the formulation of a huge mathematical model for reassigning crews
to flights when schedule disruptions occur.
 Because the model needs to be applied immediately when a disrup-
tion occurs, a decision support system called CrewSolver was devel-
oped to incorporate both the model and a huge in-memory data
store representing current operations.
 CrewSolver enables a crew coordinator to input data about the
schedule disruption and then to use a graphical user interface to re-
quest an immediate solution for how to reassign crews to flights.

HSLee@SKKU
 19/21

2.6 IMPLEMENTATION
 Implementation Steps
 First, the OR team gives operating management a careful expla-
nation of the new system to be adopted and how it relates to
operating realities.
 Next, these two parties share the responsibility for developing the
procedures required to put this system into operation.
 Operating management then sees that a detailed indoctrination is
given to the personnel involved, and the new course of action is
initiated.
 If successful, the new system may be used for years to come.
 With this in mind, the OR team monitors the initial experience
with the course of action taken and seeks to identify any modifi-
cations that should be made in the future.

HSLee@SKKU
 20/21

IMPLEMENTATION Example
 Example Case - Samsung Electronics Corp. (Textbook p.20)
 initiated a major OR study in March 1996 to develop new methodologies
and scheduling applications that would streamline the entire semicon-
ductor manufacturing process and reduce work-in-progress inventories.
 The study continued for over five years, culminating in June 2001, largely
because of the extensive effort required for the implementation phase.
 The OR team needed to gain the support of numerous managers, manu-
facturing staff, and engineering staff by training them in the principles
and logic of the new manufacturing procedures.
 Ultimately, more than 3,000 people attended training sessions.
 The new procedures then were phased in gradually to build confidence.
 However, this patient implementation process paid huge dividends. The
new procedures transformed the company from being the least efficient
manufacturer in the semiconductor industry to becoming the most effi-
cient.
 This resulted in increased revenues of over $1 billion by the time the im-
plementation of the OR study was completed.

HSLee@SKKU
1:32 / 5:26
3 주차 Discussion 주제 (3 월 18 21/21

일)
 다음 OR and Analytics 관련 Case study 논문 5 개 중 1 개를 골라 읽기
 모든 학생 : 1 개 읽은 논문 한글 요약 후 3 월 17 일 일요일 오후 10 시까지 제출 (10
point, 1.5 줄 (150%) 줄간격 A4 용지 1 쪽 분량 )
 발표 학생 5 명 : 3 월 18 일 토론 시간에 각자 선택한 1 편의 논문에 대해 요약을
PPT 10 쪽 이내로 요약

1. Merrill Lynch. Pricing Analysis for Merrill Lynch Integrated Choice

2. Continental Airlines. A New Era for Crew Recovery at Continental Airlines
3. Rijkswaterstaat. Planning the Netherlands' Water Resources
4. IBM. Optimizer: IBM's Multi-Echelon Inventory System for Managing Service
Logistics
5. Samsung Electronics Corp. SLIM - Short Cycle Time and Low Inventory in
Manufacturing at Samsung Electronics

HSLee@SKKU