BMAF 201

Quantitative Methods for Business Decisions

Topic 2: Linear Programming
1. Mathematical models: Models – Linear and non-linear models
2. Systems of mathematical models
3. Optimization of systems of mathematical models (linear and non-linear
models)
4. Linear programming / Linear optimization
5. Components of linear programming
6. Assumptions of linear programming
7. Linear programming problem (LPP)
a) Characteristics of linear programming problem
b) Formulate maximization and minimization LPP - steps
c) Methods to solve linear programming problems
i. Graphical method of solving LPP (Max. & Min. problems)
ii. Simplex method of solving LPP (Max. & Min. problems)
d) Irregular types of linear programming problem
Quiz 3 (13.11.2023) – Testing understanding on topic 1 (Introduction to
Quantitative Method for Business Decisions)

How can quantitative methods be useful for below groups of potential users
a. Monetary economist who is tasked to design the monetary policy of the
country.
b. An analyst at an investment bank.
 Mathematical Models
A model is a representation of real objects or situations. Eg. A business model which
tries to explain an end to end system of a business operations. A football model (4-4-2),
to represent how the football team will line up and play.
What about economic model?? What does it try to represent?
Given set of assumptions, a mathematical model – describe/represent a
problem/relationship by using equations or mathematical procedures. Equations will
contain the variables and parameters.

𝑦 𝛼 = 𝐴𝑥1 𝛽 + 𝐵𝑥2 𝜆 + ⋯ … . +Ω𝑥𝑛 𝜏

Where;
• A, B, Ω are parameters of the model
• 𝑦, 𝑥1 , 𝑥2 , 𝑥𝑛 are variables of the model
• 𝛼, 𝛽, 𝜆, 𝜏 are the power of the variables which determine the linearity of the
mathematical model
• 𝐴, 𝐵, Ω, 𝑦, 𝑥1 , 𝑥2 , 𝑥𝑛 , 𝛼, 𝛽, 𝜆, 𝜏≥ 0 is the assumption of the model
Linearity of the mathematical model

𝑦 𝛼 = 𝐴𝑥1 𝛽 + 𝐵𝑥2 𝜆 + ⋯ … . +Ω𝑥𝑛 𝜏

From the above mathematical model

i. Generate a linear model
ii. Generate a non – linear model
 Systems Of Mathematical Models
System of mathematical model – shows a relationship between different components
represented by mathematical models (linear and/or non-linear models) to map the
activities, actions, behaviors of economic agents such as individual, firms, and
society (whole economy).

Closed economy system (Economic activities)

A closed economy system will combine the relationship between three sectors of
consumption, investment, and government expenditure, represented by their
functions

𝐺𝐷𝑃 = 𝐶 + 𝐼 + 𝐺
𝑪 = 𝑎 + 𝑏 𝑌 − 𝑇𝑎𝑥  Consumption function
𝑰 = 𝐼 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑟𝑎𝑡𝑒, 𝑜𝑡ℎ𝑒𝑟 𝑓𝑎𝑐𝑡𝑜𝑟𝑠
𝑮 = 𝐺(𝑡𝑎𝑥 𝑝𝑜𝑙𝑖𝑐𝑦, 𝑓𝑖𝑠𝑐𝑎𝑙 𝑝𝑜𝑙𝑖𝑐𝑦, 𝑚𝑜𝑛𝑒𝑡𝑎𝑟𝑦 𝑝𝑜𝑙𝑖𝑐𝑦, 𝑜𝑡ℎ𝑒𝑟 𝑓𝑎𝑐𝑡𝑜𝑟𝑠)
System Mathematical Model
Economic system – System of mathematical models /equations to describe the relationships
Closed Economy between different economic variables such as GDP, C, I, Tax, Interest
rate
Climate system System of mathematical models that describe the relationships between
different climate variables, such as temperature, precipitation, and
greenhouse gas concentrations.
Manufacturing System of mathematical models that describe the relationships between
Firm System production and manufacturing variables, such as price, quantity, costs,
profit
Portfolio System of mathematical models that describe the relationships between
Investment System different portfolio investment variables, such as stocks, returns, risks,
stock price.
Portfolio Return = (w1 * R1) + (w2 * R2) + ... + (wn * Rn)

Other systems: Diet system  Interrelation between different nutrients and their composition
 Optimization of systems of mathematical models
Given below system of a mathematical model
𝑦 𝛼 = 𝐴𝑥1 𝛽 + 𝐵𝑥2 𝜆 + ⋯ … . +Ω𝑥𝑛 𝜏
𝑅1 =≠≤≥ a𝑥1 + 𝑏𝑥2
𝑅2 =≠≤≥ c𝑥2 + 𝑑𝑥𝑛
Where ; 𝑅1 and 𝑅2 are the constants representing constraints existing between
the variables 𝑥1 , 𝑥2 , 𝑥𝑛 such as resources constraints

Optimization of that system would be to find the optimal combination of

variables 𝑥1 , 𝑥2 , 𝑥𝑛 that would either maximize or minimize the objective of
the economic agent such as; maximizing GDP, maximizing revenues of the
firm, minimizing costs, maximizing dividend yields of the investment
portfolio.
Optimization of a mathematical model which has both the objective function
and constraint functions as linear models are referred as Linear optimization or
Linear programming.

Where as,

Optimization of a mathematical model in which either of the objective function

or constraints are non-linear models are referred as Non-Linear Optimization or
Non-Linear Programming. This will be covered in topic number 3
 Linear programming / Linear optimization
• LP / LO may be defined as the problem of maximizing or
minimizing a linear function that is subjected to linear
constraints. The constraints may be equalities or inequalities.
• The problems involve can be maximization of profit or
minimization of cost.
• The objective is to find optimal solution which can attain
highest or lowest value of the function i.e. maximize or
minimize the objective function.
 Components of linear programming

𝜋 = 𝐴𝑥1 + 𝐵𝑥2 + ⋯ … . +Ω𝑥𝑛

𝑅1 =≠≤≥ a𝑥1 + 𝑏𝑥2
𝑅2 =≠≤≥ c𝑥2 + 𝑑𝑥𝑛
𝑅3 =≠≤≥ 𝑒𝑥1 + 𝑓𝑥𝑛

The basic components of the LP are as follows:

1. Decision Variables: 𝑥1 , 𝑥2 , 𝑥𝑛
2. Constraints: 𝑅1 , 𝑅2 , 𝑅3 (Constraints functions comprise of decision variables)
3. Data: Information on decision variables and constraints represented by
constants 𝐴, 𝐵, Ω, a, b, c, d,e,f, 𝑅1 , 𝑅2 , 𝑅3
4. Objective Functions: Profit function 𝜋 comprise also the decision variables
 Assumptions of Linear Programming / Linear Optimization
When working with linear programming, the following conditions
are assumed to be true
1) The relationship between the constraints and the objective
function should be linear.
2) The number of constraints should be expressed in the quantitative
terms.
3) The linear function (i.e., objective function) is to be optimized.
 Linear Programming Problem (LPP)
• The Linear Programming Problems (LPP) is a problem that is
concerned with finding the optimal value of the given linear
function i.e. objective function. The optimal value can be either
maximum value or minimum value.
• The objective function can contain several variables, which are
subjected to the conditions and it has to satisfy the set of linear
inequalities called linear constraints.
• E.g of LPP; manufacturing problems, diet problems, transportation
problems, allocation problems etc.
 Characteristics of linear programming problem
1. Constraints – The limitations should be expressed in the
mathematical form, regarding the resource.
2. Objective Function – In a problem, the objective function should be
specified in a quantitative way.
3. Linearity – The relationship between two or more variables in the
function must be linear. It means that the degree of the variable is
one.
4. Non-negativity – The variable value should be positive or zero. It
should not be a negative value.
5. Decision Variables – The decision variable will decide the
output. It gives the ultimate solution of the problem. For any
problem, the first step is to identify the decision variables.
 Formulating linear programming problem – steps

i) Understand the problem: The goal of a linear programming problems is

to find a way to get the most, or least, of some quantity i.e. profit or
expenses. This quantity is called your objective. The answer should
depend on how much of some decision variables you choose. Your
options for how much will be limited by constraints stated in the
problem.
ii) Describe the objective: What are you trying to optimize? Are you trying
to minimize costs? Maximize production quantities?
iii) Describe the constraints: What are the limits on "how much" your
decision variables can be? Look for words like "at least", "no more than",
"two thirds of" etc. Standard constraints for max. 𝑎𝑥1 + 𝑏𝑥2 ≤ 𝑐, and
for min. 𝑎𝑥1 + 𝑏𝑥2 ≥ 𝑐 (But this is not a must)
iv) Define the decision variables: The answer to a linear programming
problem is always "how much" of some things. What are those things?
Choose variables to represent how much of each of those things. E.g.
𝑥1 = 𝑄𝑢𝑎𝑛𝑡𝑖𝑡𝑦 𝑜𝑓 𝑠𝑢𝑔𝑎𝑟 and 𝑥2 = 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑙𝑎𝑏𝑜𝑟
v) Write the objective function: Use the variables you just chose to write
down an algebraic expression that describes the amount you're trying to
minimize. Do not use an = sign. Do not use < or > signs
vi) Write the constraints in terms of the decision variables: For each
constraint such as "at least TZS 5,000" or "no more than 400" write an
inequality using the decision variables. For example: 𝑥1 > 5,000 or 𝑥1 +
𝑥2 < 400
vii) Add the non-negativity constraints: Don't forget to include non-
negativity constraints on the variables of choice i.e. 𝑥1 ≥ 0, 𝑥2 ≥ 0
viii) Write up your optimization problem:
Maximize 4𝑥1 + 2𝑥2
Subject to: 𝑥1 + 5𝑥2 ≤ 125
𝑥1 + 𝑥2 ≤ 400
𝑥1 ≥ 0, 𝑥2 ≥ 0
Tasks: Formulate Linear Programming Problems on below

Task 1: James holds two part-time jobs, Job I and Job II. He
never wants to work more than a total of 12 hours a week. He
has determined that for every hour he works at Job I, he needs
2 hours of preparation time, and for every hour he works at Job
II, he needs one hour of preparation time, and he cannot spend
more than 16 hours for preparation. If he makes TZS 4,000 an
hour at Job I, and TZS 3,000 an hour at Job II, how many
hours should he work per week at each job to maximize his
income?
Task 2: At a university, Professor Symons wishes to employ
two people, John and Mary, to grade papers for his classes.
John is a graduate student and can grade 20 papers per hour;
John earns TZS 15,000 per hour for grading papers. Mary is a
post-doctoral associate and can grade 30 papers per hour; Mary
earns TZS 25,000 per hour for grading papers. Each must be
employed at least one hour a week to justify their employment.
If Prof. Symons has at least 110 papers to be graded each
week, how many hours per week should he employ each
person to minimize the cost?
 Methods to Solve Linear Programming Problems
The linear programming problem can be solved using different
methods such as;
1. Graphical method
2. Simplex method
3. Using tools such as R programming and open solver

At this level we will cover only graphical and simplex method

Consider Task 1 in the previous session.

Maximize
𝑌 = 4000𝑥1 + 3000𝑥2
Subject to
𝑥1 + 𝑥2 ≤ 12
2𝑥1 + 𝑥2 ≤ 16
𝑥1 ≥ 0, 𝑥2 ≥ 0
𝑥2

16
Feasible region points
A = (0,12)
B = (4, 8)
C = (8,0)
12

8 𝑥1
12
Note: Maximum (and minimum) of the objective
function lies at a corner point of the feasible region.
 Maximum income in feasible region points

Combination Maximum income as per objective function

A (0,12) 4000(0) + 3000(12) = TZS 36,000
B (4,8) 4000(4) + 3000(8) = TZS 40,000
C (8,0) 4000(8) + 3000(0) = TZS 24,000

Therefore, to maximize his income, James should work 4 hours at job I

and 8 hours at job II and boing so he will be able to maximize his
income to TZS 40,000
You as an investment analyst you are presented with this case from your
client - John
John has TZS 3,000,000 available to invest in the Dar es salaam Stock
Exchange, with which he can buy shares of Twiga Cement and CRDB Bank.
A share of Twiga Cement costs TZS 3,000 and pays dividends of TZS 400 per
share each year. CRDB Bank share costs TZS 300 per share and pays
dividends of TZS 20 per share each year. John wishes to purchase no more
than 3000 total shares of stock in order to avoid paying higher trading fees to
his broker. Due to the volatility of the Twiga Cement stock, his broker
recommends that he invest in at most 600 shares of that company’s stock.
Determine the number of shares John should buy of each kind of stock in
order to maximize the amount from dividends that he can receive at the end
of the first year.
 Graphical method – Minimization problem

Consider task 2 in the previous session

Minimize 𝐶 = 15,000𝑥1 + 25,000𝑥2
Subject to
20𝑥1 + 30𝑥2 ≥ 110
𝑥1 ≥ 1
𝑥2 ≥ 1
𝑥1 ≥ 0, 𝑥2 ≥ 0
𝑥2

Feasible region points

A = (1,3)
B = (4,1)

3.67

1 5.5 𝑥1
 Minimum cost in the feasible region
Combination Cost as per objective function
A (1,3) 15,000(1) + 25,000(3) = TZS 90,000
B (4,1) 15,000(4) + 25,000(1) = TZS 85,000

Therefore, minimize cost per week, Professor Symons

should employ John for 4 hours and Mary for 1 hour. The
minimum cost per week for grading paper will be TZS
85,000
 Limitations of graphical method

1. When constraints or objective functions are not

linear, this technique cannot be used.
2. The method is inapplicable to problems having three
or more variables.
 Simplex Method
In practice, linear programs can contain thousands of
variables and constraints in which the graphical
method can not be employed to solve them, and thus
simplex algorithm will be used.

Simplex algorithm is a numerical solution method

that uses matrices and row operations.
 Simplex method – Maximization (Steps)

Step 1: Formulate linear programming problem

Consider Task 1 in the previous session.
Maximize
𝑌 = 4000𝑥1 + 3000𝑥2
Subject to
𝑥1 + 𝑥2 ≤ 12
2𝑥1 + 𝑥2 ≤ 16
𝑥1 ≥ 0, 𝑥2 ≥ 0
Step 2: Convert inequalities into equation by adding one slack variable
for each inequality and arrange the equations. (Inequalities to occupy
top rows)
−4000𝑥1 − 3000𝑥2 + 𝑌 = 0
𝑥1 + 𝑥2 + 𝑦1 = 12
2𝑥1 + 𝑥2 + 𝑦2 = 16
Step 3: Construct the initial simplex tableau

Basic
𝑥1 𝑥2 𝑦1 𝑦2 𝑌 𝑅𝐻𝑆 solutions
1 1 1 0 0 12 𝑦1
2 1 0 1 0 16 𝑦2
-4000 -3000 0 0 1 0 𝑌
Step 4: Identify pivot column that is, the column with the
most negative entry in the bottom row. (Most –ve entry
represent largest coefficient in the objective function)

Basic
𝑥1 𝑥2 𝑦1 𝑦2 𝑌 𝑅𝐻𝑆 solutions
1 1 1 0 0 12 𝑦1
2 1 0 1 0 16 𝑦2
-4000 -3000 0 0 1 0 𝑌

Pivot column
Step 5: Identify pivot row that is, the row with the smallest
quotient. (Quotient is identified by taking RHS element
divide by the element in the pivot column

Basic
𝑥1 𝑥2 𝑦1 𝑦2 𝑌 𝑅𝐻𝑆 Quotient solutions
1 1 1 0 0 12 12/1 = 12 𝑦1
2 1 0 1 0 16 16/2 = 8 𝑦2
-4000 -3000 0 0 1 0 𝑌

The intersection between pivot column and pivot row will form a
pivot element which will be used for pivoting.
Step 6: Perform pivoting on the simplex tableau. Pivoting is
the process of obtain a 1 in the location of pivot element
and then making other elements zero in that column
To make pivot element 1  Divide pivot row by 2

Basic
𝑥1 𝑥2 𝑦1 𝑦2 𝑌 𝑅𝐻𝑆 solutions
1 1 1 0 0 12 𝑦1
1 1/2 0 1/2 0 8 𝑦2
-4000 -3000 0 0 1 0 𝑌
To make element in the pivot column above pivot element
zero  Take second row times -1 plus first row

Basic
𝑥1 𝑥2 𝑦1 𝑦2 𝑌 𝑅𝐻𝑆 solutions
-1+1 -½+1 0+1 -½+0 0-0 -8+12 𝑦1
1 1/2 0 1/2 0 8 𝑦2
-4000 -3000 0 0 1 0 𝑌
Basic
𝑥1 𝑥2 𝑦1 𝑦2 𝑌 𝑅𝐻𝑆 solutions
0 ½ 1 -½ 0 4 𝑦1
1 1/2 0 1/2 0 8 𝑦2
-4000 -3000 0 0 1 0 𝑌
To make element in the pivot column below pivot element
zero  Take second row times 4000 plus the bottom row

Basic
𝑥1 𝑥2 𝑦1 𝑦2 𝑌 𝑅𝐻𝑆 solutions
0 ½ 1 -½ 0 4 𝑦1
1 ½ 0 ½ 0 8 𝑥2
0 -1000 0 2000 1 32,000 𝑌

Step 7: When there are no more negative entries in the

bottom row, we are finished, otherwise, we start from
step 4
 Basic
𝑥1 𝑥2 𝑦1 𝑦2 𝑌 𝑅𝐻𝑆 Quotient solutions
0 ½ 1 -½ 0 4 4/0.5 =8 𝑦1
1 ½ 0 ½ 0 8 8/0.5 =16 𝑥2
0 -1000 0 2000 1 32,000 𝑌

Make pivot element 1  Multiply first row by 2

Basic
𝑥1 𝑥2 𝑦1 𝑦2 𝑌 𝑅𝐻𝑆 solutions
0 1 2 -1 0 8 𝑦1
1 ½ 0 ½ 0 8 𝑥2
0 -1000 0 2000 1 32,000 𝑌
Make element below pivot element zero  Multiply first
row by -1/2 and add to the second row

Basic
𝑥1 𝑥2 𝑦1 𝑦2 𝑌 𝑅𝐻𝑆 solutions
0 1 2 -1 0 8 𝑦1
1 0 -1 1 0 4 𝑥2
0 -1000 0 2000 1 32,000 𝑌

Make bottom row element in the pivot column zero  Multiply

the first row by 1000 and add to the bottom row
𝑥1 𝑥2 𝑦1 𝑦2 𝑌 𝑅𝐻𝑆 Basic solutions
0 1 2 -1 0 8 𝑥2
1 0 -1 1 0 4 𝑥1
0 0 2000 1000 1 40,000 𝑌
Solution
𝑥1 = 4, 𝑥2 = 8, 𝑎𝑛𝑑 𝑌 = 40,000
Simplex method - minimization
Note:
• In the case of ‘ ≤ ’ constraints, slack variables are added to the
actual variables to make the equation an equality.
• In the case of ‘ ≥ ’ constraints, surplus variables are used to make
the equation an equality.
Simplex method - minimization
• For minimization problem using simplex, we will
employ a dual problem.
• To every minimization problem there corresponds a
dual problem. The solution of the dual problem is used
to find the solution of the original problem.

Duality principle - The objective function of the

minimization problem reaches its minimum if and only
if the objective function of its dual reaches its
maximum. And when they do, they are equal.
Consider task 2 in the previous session

Minimize 𝐶 = 15,000𝑥1 + 25,000𝑥2

Subject to
20𝑥1 + 30𝑥2 ≥ 110
𝑥1 ≥ 1
𝑥2 ≥ 1
𝑥1 ≥ 0, 𝑥2 ≥ 0

To solve the above LPP using simplex method, below

steps will be followed.
Step 1: Arrange the problem as a matrix in order to
convert the minimization problem into its dual
20 30 110
1 0 1
0 1 1
15,000 25,000 0

Step 2: Find the transpose of the above matrix

20 1 0 15,000
30 0 1 25,000
110 1 1 0
Step 3: Form a maximization problem from the
transpose matrix above which will be a dual problem
of the minimization problem

Maximise D = 110𝑦1 + 𝑦2 + 𝑦3
Subject to
20𝑦1 + 𝑦2 ≤ 15,000
30𝑦1 + 𝑦3 ≤ 25,000
𝑦1 ≥ 0, 𝑦2 ≥ 0, 𝑎𝑛𝑑 𝑦3 ≥ 0
Step 4: Construct initial simplex tableau from the
above problem
𝒚𝟏 𝒚𝟐 𝒚𝟑 𝒙𝟏 𝒙𝟐 𝑫 𝑹𝑯𝑺
20 1 0 1 0 0 15,000
30 0 1 0 1 0 25,000
-110 -1 -1 0 0 1 0
Step 5: Find pivot column, pivot row, and pivot
element (Just like the previous session) Note: Pivot
element cannot be zero.

𝒚𝟏 𝒚𝟐 𝒚𝟑 𝒙𝟏 𝒙𝟐 𝑫 𝑹𝑯𝑺
20 1 0 1 0 0 15,000
30 0 1 0 1 0 25,000
-110 -1 -1 0 0 1 0
Step 6: Do pivoting
• To make pivot element 1  Multiply the 1st row by
1/20
𝒚𝟏 𝒚𝟐 𝒚𝟑 𝒙𝟏 𝒙𝟐 𝑫 𝑹𝑯𝑺
1 1/20 0 1/20 0 0 750
30 0 1 0 1 0 25,000
-110 -1 -1 0 0 1 0
• To make element in 2nd row below pivot element zero
 Multiply 1st row by -30 and add to 2nd row
𝒚𝟏 𝒚𝟐 𝒚𝟑 𝒙𝟏 𝒙𝟐 𝑫 𝑹𝑯𝑺
1 1/20 0 1/20 0 0 750
0 -30/20 1 -30/20 1 0 2,500
-110 -1 -1 0 0 1 0
• To make element in the 3rd row below pivot element
zero  Multiply 1st row by 110 and add to the 3rd row

𝒚𝟏 𝒚𝟐 𝒚𝟑 𝒙𝟏 𝒙𝟐 𝑫 𝑹𝑯𝑺
1 1/20 0 1/20 0 0 750
0 -30/20 1 -30/20 1 0 2,500
• To make
0 element
4.5 rd row0below
in-1the 35.5 1 pivot82,500
element
zero  Add 2nd row to the 3rd row

𝒚𝟏 𝒚𝟐 𝒚𝟑 𝒙𝟏 𝒙𝟐 𝑫 𝑹𝑯𝑺
1 1/20 0 1/20 0 0 750
0 -30/20 1 -30/20 1 0 2,500
0 3 0 4 1 1 85,000
3.0 Irregular types of LPP
In theory, every linear programming problem either:
1. Is infeasible
2. Is unbounded
3. Has a unique optimal solution value (Single optimal
point or multiple optima)
Irregular LPP occurs under below circumstances.
1. When LPP problem has no feasible solution area.
2. LPP with multiple optimal points i.e. multiple
optima
3. Problems with unbounded solutions
4. Problems with ties for the pivot column or ties for
the pivot row.
5. Problems with constraints with negative quantity
values i.e. breach the non-negativity condition
Note:For irregular problems the general simplex
procedure does not always apply.
3.0.1 Infeasible LP models
An LLP is infeasible if it has no feasible solution in
which there will be no solution that will satisfy all
constraints. Given below LPP
Maximize 8𝑥1 + 5𝑥2
Subject to. 2𝑥1 + 𝑥2 ≤ 1000
3𝑥1 + 4𝑥2 ≤ 2400
𝑥1 − 𝑥1 ≤ 350
𝑥1 ≥ 800
𝑥1 , 𝑥2 ≥ 0
 3.0.2 Unbounded solution

An unbounded solution means that the feasible region

extends to infinity in some direction which also imply
that the objective function will have an infinity
solution. Consider below LPP problem

Maximize 8𝑥1 + 5𝑥2

Subject to. 𝑥1 − 𝑥1 ≤ 350
𝑥1 ≥ 200
𝑥1 ≥ 300
𝑥1 , 𝑥2 ≥ 0
Unbounded solution is irrelevant because it indicates that there
is possibility to have an infinity profit which is not realistic
3.0.3 Multiple optima
Happens when there are more than one optimal
points which can optimise the objective function.
Consider below LPP.

Maximize 8𝑥1 + 4𝑥2

Subject to. 2𝑥1 + 𝑥2 ≤ 1000
3𝑥1 + 4𝑥2 ≤ 2400
𝑥1 − 𝑥1 ≤ 350
𝑥1 , 𝑥2 ≥ 0
Both optimal extreme points will give the same figure for
maximum profit