2021-12-01

Operations Research
ENGR 3157
Lecture 10 – Goal Programming

Eugene Ben-Awuah, PhD, PEng

Bharti School of Engineering
Laurentian University

Office: F220
Tel: 705-675-1151 ext. 2195
Email: [email protected]

2
Goal Programming

• Definition:
– Generalization of linear programming
• Used to handle multiple conflicting objectives
– Each objective is given a “goal”
– Seeks a “compromise” solution
– Multi-criteria decision analysis

1
 2021-12-01

3
Goal Programming General Formulation

In general,
Minimize
(𝑎𝑟𝑠𝑟 + 𝑏𝑟𝑠𝑟 )

Subject to
𝑐𝑟 𝑥1, … , 𝑥𝑛 + 𝑠𝑟 − 𝑠𝑟 = 𝑔𝑟 (𝑟 = 1, … , 𝑘)
𝑠𝑟 , 𝑠𝑟 ≥ 0 𝑟 = 1, … , 𝑘
𝑥𝑖 ≥ 0 (𝑖 = 1, … , 𝑛)
𝑎𝑟, 𝑏𝑟, 𝑔𝑟 ∈ ℝ

4
Methods of Goal Programming

• Non-Pre-Emptive Method
– Also called the Weights Method

• Pre-emptive method

2
 2021-12-01

5
Non-Pre-Emptive Goal Programming

An Example:
The Leon Burnit Advertising Agency is trying to determine a TV advertising
schedule for Priceler Auto Company. Priceler has three goals:
1. Its ads should be seen by at least 40 million high-income men.
2. Its ads should be seen by at least 60 million low-income people.
3. Its ads should be seen by at least 35 million high-income women.
Leon Burnit can purchase two types of ads: those shown during football
games and those shown during soap operas. At most $600,000 can be
spent on ads. Advertising costs and potential audiences of a one-minute ad
are shown below. Determine how many of each ad to purchase for Priceler.

6
Non-Pre-Emptive Goal Programming

Let
• x1 = number of minutes of ads shown during
football games.
• x2 = number of minutes of ads shown during soap
operas.

Formulate the constraints:

3
 2021-12-01

7
Non-Pre-Emptive Goal Programming

Convert constraints to equality by introducing

deviational variables
• si+ = amount by which we exceed constraint i.
• si- = amount by which we fall short of constraint i.

8
Non-Pre-Emptive Goal Programming

We have the following goals:

• Minimize G1 = s1-
• Minimize G2 = s2-
• Minimize G3 = s3-

For both optimization methods, we form a single objective

function:
• Minimize 𝑧 = 𝑤1𝐺1 + 𝑤2𝐺2 + 𝑤3𝐺3

4
 2021-12-01

9
Pre-Emptive Method

• Hierarchy of priority levels for objectives

– One goal more important than another
• Can later divide into second, third, etc. priorities
– Lower one-sided goals
– Upper one-sided goals
– Two-sided goals

12
Non-Pre-Emptive GP: Comparison to Regular LP

• A Non-Pre-Emptive goal program is a linear program

• The main difference is that

– Linear Programming finds an optimal solution to the original problem
– Goal Programming finds a compromise solution to the original
problem

5
 2021-12-01

13
Pre-Emptive GP: Comparison to Regular LP

• Normal LP Simplex needs 1 row 0

• Goal programming Simplex needs a row 0 per goal
• When a pivot is done in goal programming, each row 0
needs updating
• Tableau is optimal when
• All goals are satisfied
• Selecting Entering Variable
• Select highest priority goal not met

15
Pros and Cons of Goal Programming

Pros Cons
• Able to use the simplex • Goal targets and weights
algorithm determination is
• Can handle multiple subjective
objectives • Cannot trade off small
• Small input requirement degradation of high
– Goal targets and weights priority objective for large
improvement to low
priority objective
• Obtain a compromise
solution

6
 2021-12-01

16
Example: Pre-Emptive GP

Suppose each million exposures by which Priceler

falls short of constraint number …
• One costs $200,000 in lost sales
• Two costs $100,000 in lost sales
• Three costs $50,000 in lost sales

17
Example: Pre-Emptive GP

7
 2021-12-01

18
Goal Programming – Mining Application

• MILGP model objective function =

Objective functions: (a) + (b) + (c)

J



 
  
 vku ,t  xku ,t  qku ,t  yku ,t  pku ,t  zku ,t  hku ,t  sku ,t    
 

U T

  Max


    P1 a1d1  P2 a2 d 2 
  ,u , t
 
 , u ,t
    
  
j 1 u 1 t 1 k B
 
  
 P a d  ,u ,t  a d  ,u ,t  P a d  ,u ,t  a d  ,u ,t     
j

   3 3 3 3 3 4 4 4 4 4   

19
Goal Programming – Mining Application

• MILGP model goal functions:

– Mining goal; Processing material goal; OI Dyke material goal; TCS
Dyke material goal

J
 
   o k
 wk  d k   yk   d1
u ,t  ,u ,t
 Tm
u ,t

j 1  k B j 
 
    ok  xku ,t    d 2 ,u ,t  Tpu ,t
J

j 1  k  B j 
 
    d k  zku ,t    d3 ,u ,t  d3 ,u ,t  Tdu ,t
J

j 1  k  B 
j

 
    lk  sku ,t    d 4 ,u ,t  d 4 ,u ,t  Tl u ,t
J

j 1  k  B j 

8
 2021-12-01

20
Large Scale Optimization
• Recall Branch and Bound
– Branch and Bound + Cutting Planes = Branch and Cut (CPLEX,
GUROBI, etc)
• Cutting Planes
– Mathematical cutting plane algorithms
• Gomory fractional cuts
– Problem structure cutting plane techniques
• Mining related
– Period aggregation
– Mining capacity/Period cuts
– Initial solution
– Other examples from other industry

• Discuss Application of Goal Programming and Large

Scale Optimization

24
References

Hillier, F. (2014). Introduction to operations research. New York: McGraw-Hill.

Ignizio, J. P. (1989). On the merits and demerits of integer goal programming. The
Journal of the Operational Research Society, 40(8), 781-785.

Min, H., & Storbeck, J. (1991). On the origin and persistence of misconceptions in
goal programming. The Journal of the Operational Research Society, 42(4), 301-
312.

Taha, H. (2010). Operations research: An introduction (9E). New Jersey: Pearson.

Winston, W. L. (2004). Operations research: Applications and algorithms (4E).

California: Thomson.

Zeleny, M. (1980). The pros and cons of goal programming. Computers and
Operations Research, 8(4), 357-359.

9
 2021-12-01

26
Branch and Bound Take Home

Max Z = 2x1  x2
Subject to: 5 x1  2 x2  8
x1  x2  3
x1, x2  0; x1 integer

10