Linear Programming

Developed by Dantzig in the late 1940s A mathematical method of allocating scarce resources to achieve a single objective The objective may be profit, cost, return on investment, sales, market share, space, time LP is used for production planning, capital budgeting, manpower scheduling, gasoline blending By companies such as AMD, New York Life, Chevron, Ford
DSCI 3023 1

Linear Programming
Decision Variables
symbols used to represent an item that can take on any value (e.g., x1=labor hours, x2=# of workers)

Parameters
known constant values that are defined for each problem (e.g., price of a unit, production capacity)

Decision Variables and Parameters will be defined for each unique problem
DSCI 3023 2

Linear Programming
A method for solving linear mathematical models Linear Functions
f(x) = 5x + 1 g(x1, x2) = x1 + x2

Non-linear functions
f(x) = 5x2 + 1 g(x1, x2) = x1x2 + x2
DSCI 3023 3

Formulating an LP
Define the decision variables identifying the key variables whose values we wish to determine Determine the objective function determine what we are trying to do Maximize profit, Minimize total cost Formulate the constraints determine the limitations of the decision variables
DSCI 3023 4

3 Parts of a Linear Program

Objective Function Constraints Non-negativity assumptions
DSCI 3023 5

3 Parts of a Linear Program

Objective Function linear relationships of decision variables describing the problems objective always consists of maximizing or minimizing some value e.g., maximize Z = profit, minimize Z = cost

DSCI 3023

3 Parts of a Linear Program

Constraints
linear relationships of decision variables representing restrictions or rules e.g., limited resources like labor or capital

Non-negativity assumptions
restricts decision variables to values greater than or equal to zero
DSCI 3023 7

Objective Function
Always a Max or Min statement
Maximize Profit Minimize Cost

A linear function of decision variables

Maximize Profit = Z = 3x1 + 5x2 Minimize Cost = Z = 6x1 - 15x2

DSCI 3023

Constraints
Constraints are restrictions on the problem
total labor hours must be less than 50 x1 must use less than 20 gallons of additive

Defined as linear relationships

total labor hours must be less than 50
x1 + x2 < 50

x1 must use less than 20 gallons of additive

x1 < 20 DSCI 3023 9

Non-negativity Assumptions
Negative decision variables are inconceivable in most LP problems Minus 10 units of production or a negative consumption make no sense The assumptions are written as follows
x1, x2 > 0

DSCI 3023

10

LP Standard Form
Each LP must contain the three parts A standard form for a LP is as follows:
Maximize subject to Z = c1x1 + c2x2 + ... + cnxn a11x1 + a12x2 + ... + a1nxn < b1 . am1x1 + am2x2 + ... + amnxn < bm x1, x2, xn > 0
11

DSCI 3023

Methods of Solving LP Problems

Two basic solution approaches of linear programming exist The graphical Method
simple, but limited to two decision variables

The simplex method

more complex, but solves multiple decision variable problems
DSCI 3023 12

Graphical Method
1. Construct an x-y coordinate plane/graph 2. Plot all constraints on the plane/graph 3. Identify the feasible region dictated by the constraints 4. Identify the optimum solution by plotting a series of objective functions over the feasible region 5. Determine the exact solution values of the decision variables and the objective function at the optimum solution
DSCI 3023 13

Some Applications of LPs + IPs

Production Planning: Given several products with varying production requirements and cost structures, determine how much of each product to produce in order to maximize profits. Scheduling: Given a staff of people, determine an optimal work schedule that maximizes worker preferences while adhering to scheduling rules. Portfolio Management: Determine bond portfolios that maximize expected return subject to constraints on risk levels and diversification.

And an incredible number more. DSCI 3023 14

Graphing 2-Dimensional LPs

Example 1:
4

y x+y x+2y2 x3 y4 x0 y0
1 3

Optimal Solution

Maximize Subject to:

Feasible Region

x
0 1 2 3

These LP animations were DSCI 3023 created by Keely Crowston.

15

Graphing 2-Dimensional LPs

Example 2:
4

y x-y

Multiple Optimal Solutions!

Minimize ** Subject to:

1/3 x + y 4
-2 x + 2 y 4 x3

Feasible Region

x0 y0

DSCI 3023

16

Graphing 2-Dimensional LPs

Example 3:
40

y x + 1/3 y x + y 20
30

Minimize Subject to:

-2 x + 5 y 150 x5 x 0 y 0
Optimal Solution

20

Feasible Region

10

x
0 0 10 20 30 40

DSCI 3023

17

Do We Notice Anything From These 3 Examples?

y
4 3 2 1 0 0 1 2 3 4 3 2 1

y
40 30 20 10

0 0 1 2 3

0 0 10 20 30 40

A Fundamental Point

y
4 3 2 1 0 0 1

y
4
3 2 1 40 30 20 10

0 0 1 2

0 0 10 20

x
30 40

If an optimal solution exists, there is always a DSCI 3023corner point optimal solution! 19

Graphing 2-Dimensional LPs

Example 1: Maximize Subject to: x+y x+2y2 x3 y4 x 0 y 0
1 3

Second Corner pt.

4

Optimal Solution

Feasible Region

Initial Corner 0 pt.

x
0 1 2 3

DSCI 3023

20