You are on page 1of 20

What is Management Science?

 A field of study that uses applied mathematics, statistics and


computers (algorithmic and computational approaches) to
solve real world problems.
 Also known as
 Operations Research
 Decision Science

 Rapid development in 1940s and 1950s (World War II)


 Landmark event: George Dantzig discovers the simplex method in 1947
 Part of Business Analytics (which in general is about making decisions
based on data)
 Related to Industrial Engineering (branch of engineering concerned
with the design and operation of industrial processes)

1
The Modeling Approach to Decision Making
 Everyone uses models to make decisions.
 Types of models:
 Mental (arranging furniture)
 Visual (blueprints, road maps)
 Physical/Scale (aerodynamics, buildings)
 Mathematical (what we’ll be studying)

2
What is a model anyway?

A model is an abstract representation of the real world


physical, social, or other systems in terms of mathematical
equations, logical relationships, computer programs, flow
diagrams, drawings, or other forms.
 

3
Principles of Modeling

1. All models are approximate; however some models are


better than others.
2. Do not build a complicated model when a simple one
would suffice.
3. Models should be validated before implementation.
4. A model should not be taken too literally (it shouldn’t
replace reality and human intuition).
5. A model cannot be better than the information that goes
into it – JI-JO (Junk in, Junk out).
What is a “Computer Model”?

5
Benefits of Modeling
 Economy - It is often less costly to analyze decision
problems using models.
 Timeliness - Models often deliver needed information
more quickly than their real-world counterparts.
 Feasibility - Models can be used to test things that would
be impossible.
 Insight - Modeling gives us understanding that improves
decision making.

6
Categories of Mathematical Models
 Prescriptive – It is assumed that certain decisions always lead
to certain outcomes. The problem is to find the best decision
(e.g. production planning).
 Decisions Xi under our control, function f known

 Descriptive – The model describing how the system works is


known, but the situations that are likely to occur are not known
(e.g. queue simulation).
 Function f is known, but possible values Xi are uncertain

 Predictive – Given past values the goal is to predict the future


value (e.g. sales forecast).
 Past values of Xi known, function f is unknown

7
Optimization (Introduction)
 Many decision problems are about how to use limited
resources such as (Constraints):
 Time
 Money
 Workers
 Oil
 Land
 Space
 Mathematical Progamming (MP) is a field of Management
Science that finds the optimal, or most efficient, way of using
limited resources to achieve the objectives of an individual of
a business; hence the name Optimization.
8
Characteristics of Optimization Problems
 All optimization problems involve:
Decision Variables – parameters under management’s control
Should be defined so as to describe all possible alternative decisions
The values of the decision variables define a policy or a plan of action
Examples: whether to support an R&D project or not – number of salespeople to hire – where
to locate a plant – production level in a given period – amount to buy from a given supplier
– price to charge for a product

Objective – a criterion to compare alternative solutions


Examples: MAX Profit, MIN Cost, MIN Risk, MAX Service Level

Constraints – describe what’s actually feasible


Operationally (e.g. production is limited by capacity of a facility)
Legally (e.g. employees may not work more than 8 hours per shift)
Logically (e.g. all parts of the budget must add up to 100%)

9
Example of a Mathematical Model
 Production planning – the decisions are:
 X1 – the number of units of product 1 to produce
 X2 – the number of units of product 2 to produce
 The objective is to maximize the total profit Y
 Y = f(X1, X2)
 There are certain constraints restricting what production
decisions are feasible (what it’s possible to do).

 In the above model X1 and X2 are independant variables


and Y is a dependant variable.
10
Example
A manufacturing run produces two types of product (1 and
2). One unit of Product 1 is sold at 4$ and Product 2 is
sold at 5$.

Decision variables:
X1 – the number of units of product 1 to produce
X2 – the number of units of product 2 to produce
The objective is to maximize the total profit Y
Y = f(X1, X2) = 4X1 + 5X2
Example

The maximum holding inventory volume is 120m3 . One


unit of Product 1 has a volume of 2m3 and one unit of
Product 2 has a volume of 3m3.

Constraint: 2X1 + 3X2 <=120


Example

A production run has to produce at least 20 items.

Constraint: X1 + X2 => 20

Plot the Feasible Region.


Constraint: 2X1 + 3X2 <=120

50

40

30

20

10

0
0 10 20 30 40 50 60
Constraint: 2X1 + 3X2 <=120
X1 + X2 => 20

50

40

30

20

10

0
0 10 20 30 40 50 60
Constraint: 2X1 + 3X2 <=120
X1 + X2 => 20

50

(0,40)
40

30
(0,20)
20

10 (60,0)
(20,0)
0
0 10 20 30 40 50 60
Constraint: 2X1 + 3X2 <=120
X1 + X2 => 20

50 Y = f(X1, X2) = 4X1 + 5X2


(0,40), Y= 200
40 Y*=240
X1*=60, X2*=0
30
(0,20),
20 Y=100

10 (60,0), Y=240
(20,0), Y=80
0
0 10 20 30 40 50 60
Overview of the Course Topics
Software in the Course

 Microsoft Excel
 Solver Add-in
• Built in, for LP, MILP, and NLP, max 200 variables and 100 constraints
 OpenSolver Add-in
• Free, more reliable, available from opensolver.org, for LP and MILP
 Palisade DecisionTools Suite
• Commercial, www.palisade.com, 15-day trial version available
• PrecisionTree for Decision Analysis, @Risk for Simulation
 Python
 Gurobi
• Commercial, state-of-the-art, gurobi.com, free student license available
• For LP, MILP, and QP (convex quadratic programming)
 scipy.optimize package for NLP
Course Philosophy: Learning by Doing

 Do’s:
 Creating your own models and programs
 Making a mistake in your model or program and fixing it
 Helping a friend understand a mistake in his/her model/code
 Comparing your model with your friend’s model

 Don’ts:
 Studying by only watching someone else create and solve models (gives a
false feeling of mastery)
 Sharing solutions of problems (quick and easy, but deprives the other
student of the valuable learning experience)
 For group projects, having only one group member do all modeling and
coding

You might also like