CHAPTER 5
Modeling and Analysis
1
5.1 Modeling and Analysis
Opening Vignette:
Siemens Solar Industries (SSI), is the world’s largest-volume
maker of solar electric products. SSI operates in an
extremely competitive market. Before 1994, the company
suffered continuous problems in photocell fabrication,
including poor material flow, unbalanced resource use,
bottlenecks in throughput, and schedule delay. To
overcome the problems, the company decided to build a
cleanroom contamination-control technology. Cleanroom
are standard practice in semiconductor business, but they
had never been used in the solar industry. The new
technology in which there is perfect control of
temperature, pressure, humidity, and air cleanliness, was
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-2
5.1 Modeling and Analysis
Shown in research to improve quality considerably. In
addition, productivity is improved because of fewer
defects, better material flow, and reduced cycle times.
Because no one in the solar industry had ever used a
cleanroom, the company decided to use a simulation,
which provided a virtual laboratory where the engineers
could experiment with various configurations of layouts
and processes before the physical systems were
constructed. Changes can be made quickly and
expensively in a simulated world because physical
changes need not be made. The simulation model enabled
the prediction of the effects of changes and what-if
analyses. A major benefit of the simulation-modeling
process was the knowledge and insight the company
gained in understanding the interactions of the systems
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-3
5.1 Modeling and Analysis
Being designed.
Computer simulation allowed SSI to compare numerous
alternatives quickly. The company attempted to find the
best design for the cleanroom and evaluate alternatives
scheduling, delivery rules, and material flow with respect
to queue (waiting line) levels, throughout, cycle time,
machine utilization, and work-in-progress levels.
The simulation was constructed with a tool called ProModel
(from ProModel Corp. Orem, UT,
http://www.promodel.com). The tool allowed the company
constructs simulation models easily and quickly and to
conduct what-if analyses. It also included extensive
graphics and animation capabilities.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-4
5.1 Modeling and Analysis
The simulation involved the entire business process, the
machines, equipment, workstations, storage and handling
devices, operators, and material and information flows
necessary to support the process. Many scenarios were
developed and experiment run. Using brainstorming, the
builders came up with many innovative suggestions that were
checked by the simulation. Incidentally, the company
involved a group of students from California Polytechnic
University in San Louis Obispo in the design and
implementation of the system.
The solution identified the best configurations for the
cleanroom, designed schedule with minimum interruptions
and bottlenecks, and improved material flow, while reducing
work-in-progress inventory levels to a minimum.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-5
5.1 Modeling and Analysis
All in all, the simulation enabled the company to improve the
manufacturing process of different solar products
significantly. The cleanroom facility has saved SSI over $75
million each year. The simulation showed how to integrate
the cleanroom with manufacturing processes in the most
efficient manner.
What have we learned from this vignette?
A complex decision
Simulation approach is used
Commercial software vs. self-developing
Save money
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-6
5.1 Modeling and Analysis
 Model is a Major DSS component, particular in Model
base-DSS and model management
 CAUTION - Difficult Topic Ahead
– Familiarity with major ideas
– Basic concepts and definitions
– Tool--influence diagram
– Model directly in spreadsheets
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-7
5.1 Modeling and Analysis
Structure of some successful models and
methodologies
Decision analysis
Decision trees
Optimization
Heuristic programming
Simulation
New developments in modeling tools / techniques
Important issues in model base management
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-8
5.1 Modeling and Analysis
 Modeling for MSS
•
•
•
•
•
•
•
•
•
•
•
•
Static and dynamic models
Treating certainty, uncertainty, and risk
Influence diagrams
MSS modeling in spreadsheets
Decision analysis of a few alternatives (decision tables and trees)
Optimization via mathematical programming
Heuristic programming
Simulation
Multidimensional modeling -OLAP
Visual interactive modeling and visual interactive simulation
Quantitative software packages - OLAP
Model base management
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-9
5.2 Modeling for MSS
 Modeling
is a Key element in most DSS :
is a Necessity in a model-based DSS
can lead to massive cost reduction / revenue increases
 Good Examples of MSS Models
– Siemens Solar Industries simulation model (opening
vignette)
– Procter & Gamble optimization supply chain
restructuring models
– Scott Homes AHP select a supplier model
– IMERYS optimization clay production model
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-10
5.2 Modeling for MSS
Major Modeling Issues








Problem identification
Environmental analysis
Variable identification
Forecasting
Multiple model use
Model categories or selection
Model management
Knowledge-based modeling
We will discuss these topics in detail.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-11
5.2 Modeling for MSS
 Problem identification and Environmental analysis
Environmental scanning and analysis, which is the
monitoring, scanning, and interpretation of the collected
information. It is often advisable to analyze the scope of the
domain and the forces and dynamics of the environment. It
is necessary to identify the organizational culture and the
corporate decision-making process (who makes decisions,
degree of centralization, and so on)
 Variable identification
The identification of the model’s variables (decision and
other) is of utmost importance, and so are their relationships.
Influence Diagrams is a useful tool. See later in section 5.5.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-12
5.2 Modeling for MSS
 Forecasting
Is essential for the construction and manipulation of the
models because the results of decision based on a model
will usually occur in the future.
 Multiple model: DSS may include several models
(sometimes dozens, each of which represents different
parts of the decision-making problem).
 Model categories or selection (Table 5.1): Table
classifies DSS models into seven groups. It also lists
several representative techniques in each category and
indicates where we discuss each one. Each technique may
be applied to either a static or a
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-13
5.2 Modeling for MSS
dynamic model, which may be constructed under assumed
environments of certainty, uncertainty, or risk. To expedite
model construction, one can use modeling language.
 Model management: To maintain their integrity and thus
their applicability, models like data, must be managed. Such
management is done with the aid of model base management
software.
 Knowledge-based modeling: DSS uses mostly quantitative
models, whereas expert systems use qualitative knowledgebased models in their application.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-14
Table 5.1 Category of models
category
Process and Objective
Representative Technique
Optimization of problems with Find the best solution from a
few alternatives
small number of alternatives
Optimization via Algorithm
Find the best solution from a
large or an infinite number of
alternatives using a step-bystep improvement process
Decision tables, decision
tree
Linear and other
mathematical programming
models, network models
Optimization via analytical
formula
Find the best solution in one
step, using a formula
Some inventory models
Simulation
Find a good enough solution, several types of simulation
or best among the alternatives
checked, using
experimentation
Heuristics
Find a good enough solution
using rules
Other models
Find a what-if using a formula Financial modeling, waiting
line
predict future for a given
forecasting models,
scenario
Markov analysis.
Predictive Models
Fall, 2006
heurisitics programming,
expert systems
All right reserved, YAO Zhong, School of E&M, BUAA
5-15
5.3 Static and Dynamic Models
 Static Analysis
– Static models take a Single snapshot of a situation. During
this snapshot everything occurs in a single interval. For
example, a decision on whether to make or to buy a
product is static in nature. A quarterly or monthly incomes
statement is static, and so is the investment decision.
– During a static analysis, stability of the relevant data is
assumed.
 Dynamic Analysis
– Dynamic models are used evaluate scenarios that changed
over time. E.g., a 5-year profits and loss projection in
which the input data, such as costs, prices, and quantities,
changed from year to year.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-16
5.3 Static and Dynamic Models
Time dependent
For example, in determining how many checkout points should be
open in a supermarket, it is necessary to consider the time of day
because there are changes in the number of people that arrive at
different hours.
Show trends and patterns over time, also show
averages per period, moving average, and comparative
analysis.
For example, profit this quarter against profits in the same quarter of
last year.
Extend static models into dynamic nature of the
problem
For example, transportation model can be extended into a dynamic
network flow model to accommodate inventory and backordering.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-17
5.4 Treating Certainty, Uncertainty,
and Risk
We have introduced these concepts at chapter 2, here we
emphasize modeling issues for each situations:
 Certainty Models
Easy to work with and yield a optimal solution.
Many financial models are constructed under assumed
certainty.
 Uncertainty
Managers attempt to avoid uncertainty as much as possible.
However, if you are no enough information, you must treat
them as an uncertain problems.
 Risk
Most business decisions are made under assumed risk.
Several techniques can be used to deal with risk analysis.
Section 5.7 will discuss in details.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-18
5.5 Influence Diagrams
 Influence Diagrams is a graphical representations of a
model used to assist in model, design, development, and
understanding.
It can provide:
Visual communication to the model builders or
development teams.
Some packages create and solve the mathematical model
Framework for expressing MSS model relationships
Rectangle = a decision variable;
Circle = uncontrollable or intermediate variable;
Oval = result (outcome) variable: intermediate or final
variables connected with arrows;
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-19
5.5 Influence Diagrams
Example, consider the following profit model:
Profit = Income – Expenses
Income= Units sold  Unit price
Units Sold = 0.5  Amounts used in advertisement
Expenses = Unit cost  Units sold + Fixed cost
Then, a simple influence diagram can be drawn:
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-20
5.5 Influence Diagrams
Figure 5.1
Fixed Cost
Expenses
Unit Cost
Profit
~
Amount used in
advertisement
Unit sold
Income
Unit Price
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-21
5.5 Influence Diagrams
The variables are connected with arrows, which indicate
that direction of the influence. The shape of the arrows
also indicates the type of relationship. For example,
Expenses
Certainty
Amount
in CDs
Interest
Collected
Profit
Uncertainty
Risk
Fall, 2006
Unit sold
Price
~
Demand
Sales
Sales
All right reserved, YAO Zhong, School of E&M, BUAA
5-22
5.5 Influence Diagrams
 Random (risk) variable : place a tilde (~) above the
variable’s name.
 Preference (usually between outcome variables) : a doubleline arrow 
 Arrows can be one-way or two-way(bidirectional),
depending on the direction of influence of a pair of
variables.
 Can be constructed at any degree of detail and
sophistication.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-23
5.5 Influence Diagrams
There are some software to create a Influence
Diagram, such as
Analytica (Lumina Decision System, Los Altos, CA,
http://www.lumina.com). Analytica supports hierarchical
diagram, multidimensional arrays, integrated
documentation, and parameter analysis.
DPL (from Applied Decision Analysis, Menlo Park, CA)
DS Lab (from DS Group Inc. Greenwich,CT)
INDIA (From Decision focus Inc. Palo Alto, CA,
http://www.dfi.com).
Precision Tree (from Palisade Corp. Newfield, NY,
http://www.palisade.com). this software can directly
creates influence diagram and decision tree in the Excel
spreadsheet.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-24
5.6 MSS Modeling in Spreadsheets
 Spreadsheet: most popular end-user modeling tool
 Powerful functions
 Add-in functions and solvers
 Important for analysis, planning, modeling
 Programmability (macros)
 Software:
@Risk (Winston Corp)
What best! (Lindo system)
Solver (Frontline Systems Inc. Now integrated in
MS Excel)
Lotus 1-2-3 (Lotus Inc.)
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-25
5.7 Decision Analysis of Few Alternatives
(Decision Tables and Trees)
Decision situations that involve a finite and usually not too
large number of alternatives are modeled by an approach
called decision analysis, in which the alternatives are
listed with their forecasted contributions to the goals, and
the probability of realizing such as a contribution, in a
table or graph. They can be evaluated to selected the best
alternative.
There are two distinct cases: a single goal and multiple goals.
Single-goal situations can be approached by using decision
tables or decision trees. Multiple goals (criteria) can be
approached by several other techniques.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-26
5.7 Decision Analysis of Few Alternatives
(Decision Tables and Trees)
Decision tables are a convenient way to organize information
in a systematic manner. For example, an investment
company is considering investing in one of three
alternatives: bonds, stocks, or certificates of deposit (CDs).
they interested in one goal : maximize the yield after one
year. (if it were interested in other goals such as safety or
liquidity, then problem would be classified as one of
multicriteria decision analysis)
 Yield depends on the status of the economy (the state of
nature) such as
– Solid growth
– Stagnation
– Inflation
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-27
5.7 Decision Analysis of Few Alternatives
(Decision Tables and Trees)
Possible Situations
1. If solid growth in the economy, bonds yield 12%;
stocks 15%; time deposits 6.5%
2. If stagnation, bonds yield 6%; stocks 3%; time
deposits 6.5%
3. If inflation, bonds yield 3%; stocks lose 2%; time
deposits yield 6.5%
Note: investment combination must also be
considered in practice. For example, 50% Stock,
50% bonds)
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-28
5.7 Decision Analysis of Few Alternatives
(Decision Tables and Trees)
View Problem as a Two-Person Game
Investment decision-making problem may be viewed as a twoperson game. The investor makes a choice (a Move) and
then nature happens (make a move). The payoff is shown in
a table representation that represents a mathematical model.
According to our definition in Chapter 2, the table includes
decision variables (alternatives), uncontrollable variables
(the states of economy), and the result variables (projected
yield). Note that all of the them structured in a spreadsheet.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-29
5.7 Decision Analysis of Few Alternatives
(Decision Tables and Trees)
View Problem as a Two-Person Game
Payoff Table 5.2
Investment problem decision table model
State of Nature (uncontrollable variables)
AlternativesSolid Growth Stagnation
Inflation
Bonds
12.00%
6.00%
3.00%
Stocks
15.00
3.00
-2.00
CDs
6.50
6.50
6.50
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-30
5.7 Decision Analysis of Few Alternatives
(Decision Tables and Trees)
Treating Uncertainty: Several approaches can handle
uncertainty.
 Optimistic approach: involves considering the best
possible outcome of each alternatives and selecting the
best of the bests (stock).
 Pessimistic approach: involves considering the worst
possible outcome for each alternative and select the best
one (CDs)
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-31
5.7 Decision Analysis of Few Alternatives
(Treating Risk)
 Use known probabilities (Table 5.3)
 Risk analysis: compute expected values
EV=(probabilityResults)
Can be dangerous: for example, Financial advisor presents
you with an “almost sure” investment of $1,000 that will
double your money in one day. Then he says, “well, there is
a 0.999 probability that you will double your money, but
unfortunatively there is a 0.0001 probability that you would
be liable for a $500,000 out-of-pocket loss”. The expected
value of this investment is
0.9999($2,000-$1,000)+0.0001(-$500,000-$1,000)=$949.8
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-32
Table 5.3: Decision Under Risk and Its Solution
Solid
Stagnation
Growth
Inflation
Expected
Value
Alternatives
.5
.3
.2
Bonds
12%
6%
3%
Stocks
15%
3%
-2%
8.0%
CDs
6.5%
6.5%
6.5%
6.5%
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
8.4% *
5-33
 Decision Trees
 Other methods of treating risk
 Simulation
 Certainty factors
 Fuzzy logic
 Multiple goals
AHP approach: Saaty: Analytic Hierarchy Process
 Yield, safety, and liquidity (Table 5.4)
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-34
Table 5.4: Multiple Goals
Alternatives Yield
Safety
Liquidity
Bonds
8.4%
High
High
Stocks
8.0%
Low
High
CDs
6.5%
Very High
High
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-35
5.8 Optimization via Mathematical
Programming
 Linear programming (LP)
Used extensively in DSS
 Mathematical Programming
Family of tools to solve managerial problems in
allocating scarce resources among various
activities to optimize a measurable goal
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-36
LP Allocation
Problem Characteristics
1. Limited quantity of economic resources
2. Resources are used in the production of products or
services
3. Two or more ways (solutions, programs) to use the
resources
4. Each activity (product or service) yields a return in
terms of the goal
5. Allocation is usually restricted by constraints
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-37
LP Allocation Model
 Rational economic assumptions
1. Returns from allocations can be compared in a common
unit
2. Independent returns
3. Total return is the sum of different activities’ returns
4. All data are known with certainty
5. The resources are to be used in the most economical manner
 Optimal solution: the best, found algorithmically
Allocation problem typically have a number of possible
alternative solutions. Depending on the underlying
assumptions, the number of solutions can either be infinite or
finite. Of the available solutions, one (sometimes more than
one) is best. In the sense that the degree of goal attainment
associated with it is the highest (total reward is maximized).
This is called an optimal solution, which can be found by
Fall,
2006
All right reserved, YAO Zhong, School of E&M, BUAA
algorithm.
5-38
Linear Programming
Decision variables
Objective function
Objective function coefficients
Constraints
Capacities
Input-output (technology) coefficients
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-39
Lindo LP Product-Mix Model
DSS in Focus 5.4
<< The Lindo Model: >>
MAX
8000 X1 + 12000 X2
SUBJECT TO
LABOR)
300 X1 + 500 X2 <=
200000
BUDGET)
10000 X1 + 15000 X2 <=
8000000
MARKET1)
X1 >=
100
MARKET2)
X2 >=
200
END
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-40
<< Generated Solution Report >>
LP OPTIMUM FOUND AT STEP
3
OBJECTIVE FUNCTION VALUE
1)
VARIABLE
X1
X2
Fall, 2006
5066667.00
VALUE
333.333300
200.000000
REDUCED COST
.000000
.000000
All right reserved, YAO Zhong, School of E&M, BUAA
5-41
ROW
LABOR)
BUDGET)
MARKET1)
MARKET2)
SLACK OR SURPLUS
.000000
1666667.000000
233.333300
.000000
NO. ITERATIONS=
Fall, 2006
DUAL PRICES
26.666670
.000000
.000000
-1333.333000
3
All right reserved, YAO Zhong, School of E&M, BUAA
5-42
RANGES IN WHICH THE BASIS IS UNCHANGED:
VARIABLE
X1
X2
OBJ COEFFICIENT RANGES
CURRENT
ALLOWABLE
ALLOWABLE
COEF
INCREASE
DECREASE
8000.000
INFINITY
799.9998
12000.000
1333.333
INFINITY
RIGHTHAND SIDE RANGES
ROW
CURRENT
RHS
LABOR
200000.000
BUDGET 8000000.000
MARKET1
100.000
MARKET2
200.000
Fall, 2006
ALLOWABLE
INCREASE
50000.000
INFINITY
233.333
140.000
ALLOWABLE
DECREASE
70000.000
1666667.000
INFINITY
200.000
All right reserved, YAO Zhong, School of E&M, BUAA
5-43
Lingo LP Product-Mix Model
DSS in Focus 5.5
<< The Model >>>
MODEL:
! The Product-Mix Example;
SETS:
COMPUTERS /CC7, CC8/ : PROFIT, QUANTITY, MARKETLIM ;
RESOURCES /LABOR, BUDGET/ : AVAILABLE ;
RESBYCOMP(RESOURCES, COMPUTERS) : UNITCONSUMPTION ;
ENDSETS
DATA:
PROFIT MARKETLIM =
8000, 100,
12000, 200;
AVAILABLE = 200000, 8000000 ;
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-44
UNITCONSUMPTION =
300, 500,
10000, 15000 ;
ENDDATA
MAX = @SUM(COMPUTERS: PROFIT * QUANTITY) ;
@FOR( RESOURCES( I):
@SUM( COMPUTERS( J):
UNITCONSUMPTION( I,J) * QUANTITY(J)) <=
AVAILABLE( I));
@FOR( COMPUTERS( J):
QUANTITY(J) >= MARKETLIM( J));
! Alternative
@FOR( COMPUTERS( J):
@BND(MARKETLIM(J), QUANTITY(J),1000000));
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-45
<< (Partial ) Solution Report >>
Global optimal solution found at step:
Objective value:
5066667.
Variable
PROFIT( CC7)
PROFIT( CC8)
QUANTITY( CC7)
QUANTITY( CC8)
MARKETLIM( CC7)
MARKETLIM( CC8)
AVAILABLE( LABOR)
AVAILABLE( BUDGET)
Fall, 2006
Value
8000.000
12000.00
333.3333
200.0000
100.0000
200.0000
200000.0
8000000.
2
Reduced Cost
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
All right reserved, YAO Zhong, School of E&M, BUAA
5-46
UNITCONSUMPTION(
UNITCONSUMPTION(
UNITCONSUMPTION(
UNITCONSUMPTION(
Row
1
2
3
4
5
Fall, 2006
LABOR, CC7)
LABOR, CC8)
BUDGET, CC7)
BUDGET, CC8)
Slack or Surplus
5066667.
0.0000000
1666667.
233.3333
0.0000000
300.00
500.00
10000.
15000.
0.00
0.00
0.00
0.00
Dual Price
1.000000
26.66667
0.0000000
0.0000000
-1333.333
All right reserved, YAO Zhong, School of E&M, BUAA
5-47
5.9 Heuristic Programming
 The determination of optimal solutions to some complex
decision problems could involves a prohibitive amount of
time and cost or may be impossible. Alternatively, the
simulation approach may be lengthy, complex, and even
inaccurate. In such situations, it is sometimes possible to
arrive at satisfying solutions more quickly and less
expensively by using heuristics.
 Heuristic procedure can be described as finding rules that
help to solve complex problems (or intermediate subproblems to discover how to set up these sub-problem for
final solution by finding the most promising paths in the
search for solutions), finding ways to retrieve and interpret
information on each experience, and then finding the methods
that lead to a computational algorithm or general solution.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-48
5.9 Heuristic Programming
 Heuristic programming (definition)is the approach of using
heuristics to arrive at feasible and “good enough” solutions to
some complex problems. “good enough” is usually in the
range of 90-99.9 percent of the objective value of an optimal
solution.
 Heuristics can be
– Quantitative
– Qualitative (in ES)
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-49
5.9 Heuristic Programming
 Methodology
Heuristic thinking does not necessarily proceed in a direct
manner. It involves searching, learning, evaluating, judging,
and then again searching, relearning, reappraisal as exploring
and probing take place. The knowledge gained from success
or failure at some point is feedback and modifies the search
process. More often than not, it is necessary to redefine either
objective or the problem, or solve related or simplified
problems before the primary one can be solved.
Tabu search heuristics are based on intelligent search strategies
to reduce the search for high-quality solutions in computer
problem solving. Essentially, the methods “remembers” what
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-50
5.9 Heuristic Programming
what high-quality and low-quality solutions it has found, and
tries to move toward other high-quality solution and away
from the low-quality ones.
Genetic algorithms: (also called evolutionary algorithms the
simplest case start with a set of randomly generated solutions
and recombine pairs of them at random to produce offspring
(the recombination into a new generation is modeled after the
process of evolution). Only the best offspring and parents are
kept to produce the next generation. Random mutations may
also be introduced.
For example,
Play a game: you and your opponent, the opponent secretly
writes down a string of six digits, such as, 001010. You guess
this number as quickly as possible.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-51
5.9 Heuristic Programming
Rules: each time you have guessed, the opponent tells you
that how many digits (but not which one) that you guess are
correct. For example, if your guess is 110101, there is no
correct digit, then score=0; if your guess is 111101, then the
score will be 1 because of the third digit you have guessed
correctly.
How many guess you are sure to find the answer? 2 power
6=64. Average you need guess 32 times. Do you need guess
all of these combination? Of course not.
Step 1: random choice 4-6 string as your initial choice.
(A) 110100
(score =1)
(B) 111101
(Score=1)
(C) 011011
(Score=4)
(D) 101100
(Score=3)
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-52
5.9 Heuristic Programming
Step 2: delete the low scores string of (A) and (B). Call (C) and
(D) parents.
Step3: Mate the parents genes by splitting each number as
shown:
(C) 01:1011
(D) 10:1100
Step 4: crossover: with first two digits of (C) crossover the last
four digits of (D), results first offspring:
(E) 011100; score =3
Similarly, (F) 101011; Score=4
It seems no any improvement of scores. Right?
Step 5: Copy the original (C) and (D) and repeat step 3 and 4,
note that this time we choice different splitting digits as the
mate genes.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-53
5.9 Heuristic Programming
we may get new offspring (G) and (H)
(C) 0110:11
(D) 1011:00
(G) 011000; score=4
(H) 1011:11;score =3
Next, select the best “couple” from all the previous solution to reproduce.
How many?
Now we select (G) and (F), duplicate and crossover. Here are the results:
(F) 1: 01011
(G) 0:11000
(I) 111000; Score =3
(J) 001011; Score =5
Also you may generate more offspring:
(F) 101:011
(G) 011:000
(K) 101000; Score =4
(L) 011011; Score =4
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-54
5.9 Heuristic Programming
Now repeat the processes with (J) and (K) as parents, duplicate
the crossover:
(J) 01101:1
(K)10100:0
(M) 001010: score =6.
This is it! You reached the solution after 13 guesses.
Each candidate solution is called a chromosome.
Reproduction, produce new generation of improved solutions
by selecting parents with higher fitness ratings or by giving
such parents greater probability to be contributors.
Crossover, use strings of binary symbols to represent solution.
Crossover means choosing a random position on the string
and exchanging the segments either to the right or to the left
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-55
5.9 Heuristic Programming
of this point with another string partitioned similarly.
Mutation: this operator was not shown in the game. It is an
arbitrary change in a situation. Sometimes it is needed to
keep the algorithm from getting stuck. The procedure
changes a 1 to 0 or 0 to 1 instead of duplicating them.
However, this change occurs with a very low probability (say,
1/1000).
Stimulating annealing.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-56
5.9 Heuristic programming
When to Use Heuristics
– Inexact or limited input data
– Both ill-structured and well-structured problem
– Complex reality, optimization is no use
– Reliable, exact algorithm not available
– Computation time excessive
– To improve the efficiency of optimization
– To solve complex problems
– For symbolic processing
– For making quick decisions
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-57
5.9 Heuristic Programming
Advantages of Heuristics
1. Simple to understand: easier to implement and explain
2. Help train people to be creative
3. Save formulation time
4. Save programming and storage on computers
5. Save computational time
6. Frequently produce multiple acceptable solutions
7. Possible to develop a solution quality measure
8. Can incorporate intelligent search
9. Can solve very complex models
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-58
5.9 Heuristic Programming
Limitations of Heuristics
1. Cannot guarantee an optimal solution
2. There may be too many exceptions to rules
3. Sequential decisions might not anticipate future
consequences
4. Interdependencies of subsystems can influence the whole
system

Heuristics successfully applied to vehicle routing
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-59
5.10 Simulation
 In MSS, Simulation means a technique for conducting
experiment (such as what-if analysis) with a digital
computer on a model of a management system
 Frequently used DSS tool: because DSS deals with semistructured or unstructured situations, it involves complex
reality, which may not be easily represented by optimization
or other models, but can often be handled by simulation.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-60
5.10 Simulation
Major Characteristics of Simulation
 Imitates reality and capture its richness (therefore, it is
not a model)
 Technique for conducting experiments
 Descriptive, not normative tool
 Often to solve very complex, risky problems
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-61
5.10 Simulation
Advantages of Simulation
1. Theory is straightforward
2. Time compression can be obtained
3. Descriptive, not normative
4. MSS builder interfaces with manager to gain intimate
knowledge of the problem
5. Model is built from the manager's perspective
6. Manager needs no generalized understanding. Each
component represents a real problem component
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-62
5.10 Simulation
7. Wide variation in problem types
8. Can experiment with different variables
9. Allows for real-life problem complexities
10. Easy to obtain many performance measures directly
11. Frequently the only DSS modeling tool for nonstructured problems
12. Monte Carlo add-in spreadsheet packages (@Risk)
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-63
5.10 Simulation
Limitations of Simulation
1. Cannot guarantee an optimal solution
2. Slow and costly construction process
3. Cannot transfer solutions and inferences to solve other
problems
4. So easy to sell to managers, may miss analytical solutions
5. Software is not so user friendly
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-64
5.10 Simulation
Simulation Methodology
Model real system and conduct repetitive experiments
1. Define problem:
The real-world problem is examined and classified. Here
we specify why simulation is necessary. The system’s
boundaries and other such aspects of problem clarification
are handled here.
2. Construct simulation model
This step involves the determination of the variables and
their relationship and the gathering of necessary data.
Often, a flowchart is used to describe the process. Then a
computer program is written.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-65
5.10 Simulation
3. Test and validate model
The simulation model must properly represent the system
under study. This is ensured by testing and validating.
4. Design experiments
Once the model has been proven valid, an experiment is
designed. Determining how long to run the simulation is
included in this step. There are two important and
conflicting objectives: accuracy and cost. It is also prudent
to identify typical, best—case and worst-case scenarios.
These help establish the ranges (of the decision variables)
in which to work and also assist in debugging the
simulation model.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-66
5.10 Simulation
5. Conduct experiments
Conducting the experiment involves issues ranging from
random number generation to presentation of the results.
6. Evaluate results
Here, we determine the meaning of the results. In addition
to statistical tools, we may use sensitivity analysis.
7. Implement solution
The implementation of simulation results involves the
same issues as any other implementation. However, the
chances of implementation are better because manager is
usually involved in the simulation process than with other
model.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-67
5.10 Simulation
Simulation Types
 Probabilistic Simulation
– Discrete distributions
– Continuous distributions
– Probabilistic simulation via Monte Carlo technique
– Time dependent versus time independent simulation
– Simulation software
– Visual simulation
– Object-oriented simulation (developing simulation
software)
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-68
5.11 Multidimensional Modeling
Performed in online analytical processing
(OLAP)
From a spreadsheet and analysis perspective
2-D to 3-D to multiple-D
Multidimensional modeling tools: 16-D +
Multidimensional modeling - OLAP (Figure
5.6)
Tool can compare, rotate, and slice and dice
corporate data across different management
viewpoints
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-69
Entire Data Cube from a Query in
PowerPlay (Figure 5.6a)
(Courtesy Cognos Inc.)
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-70
Graphical Display of the Screen
in Figure 5.6a (Figure 5.6b)
(Courtesy Cognos Inc.)
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-71
Environmental Line of Products by Drilling
Down (Figure 5.6c)
(Courtesy Cognos Inc.)
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-72
Drilled Deep into the Data: Current Month,
Water Purifiers, Only in North America
(Figure 5.6d)
(Courtesy Cognos Inc.)
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-73
5.12 Visual Spreadsheets
 User can visualize models and formulas with
influence diagrams
 Not cells--symbolic elements like influence diagram
 The software enables the user to conduct Englishlike modeling.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-74
5.13 Visual Interactive Modeling (VIM)
and Visual Interactive Simulation (VIS)
 Visual interactive modeling (VIM) Also called
 Visual interactive problem solving
 Visual interactive modeling
 Visual interactive simulation
 Use computer graphics to present the impact of different
management decisions.
 Can integrate with GIS
 Users perform sensitivity analysis
 Static or a dynamic (animation) systems (Figure 5.7)
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-75
Generated Image of Traffic at an Intersection from the
Orca Visual Simulation Environment (Figure 5.7)
(Courtesy Orca Computer, Inc.)
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-76
5.13 Conventional Simulation
 Simulation has long been established as a useful method of
giving insight into complex MSS problem. However, the
technique of simulation does not usually allow decision
makers to see how a solution to a complex problem is
developing through time, nor does it give them the ability to
interact with it. The simulation technique gives only statistical
answers at the end of a set of particular experiments. As a
result, decision makers are not an integral part of the
simulation development, and their experience and judgment
cannot be used to directly assist the study. Thus, any
conclusions obtained by the model must be taken on trust. If
the conclusions do not agree with the intuition or practical
judgment of the decision makers, a confidence gap will appear
regarding the use of the model.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-77
5.13 Visual Interactive Simulation
(VIS)
 The basic philosophy of VIS is that decision makers can
interact with the simulation model and watch the results
develop over time. This is achieved by using a visual display
unit. Decision makers can also contribute to the validation of
that model. They will have more confidence in its use
because of their own participation. They are also in a position
to use their knowledge and experience to interact with the
model to explore alternative strategies.
 Simulation can be interactive at the design stage, at the model
running stage, or both. To obtain insight into how systems
operate under different conditions, it is important to be able
interact with the model while it is running so that alternative
suggestions or directives can be tested.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-78
5.13 -Visual Interactive Simulation (VIS)
 Visual interactive models and DSS
– VIM was used with DSS in several operation management
decisions. The method consists of priming a visual
interactive model of a plant with its current status. The
model then rapidly runs on a computer, allowing
management to observe how a plant is likely to operate in
the future. A similar approach was used to assist in the
consensus negotiation among senior managers for the
development of their budget plans.
– Queuing: a DSS in such a case usually computes several
measure of performance. (such as waiting time in the
system) for the various decision alternatives. Complex
waiting line problems require simulation. The VIM can
display the size of waiting line as it changes during the
simulation runs. The VIM can also graphically present the
answer to what-if questions regarding changes in input
variables.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-79
5.14 Quantitative Software PackagesOLAP
 Preprogrammed models can expedite DSS programming
time
 Some models are building blocks of other models
– Statistical packages (SAS, SPSS, MiniTab, Systat)
– Management science packages:OSL, CPLEX, GPSS,
eXpress.
– Revenue (yield) management
– Other specific DSS applications
including spreadsheet add-ins (What’s Best, Lindo
System Inc.), Solver (Frontline System Inc.) @Risk,
@Brain,Evolver(GA add-in for Excel, etc.).
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-80
5.15 Model Base Management
 MBMS: capabilities similar to that of DBMS
 But, there are no comprehensive model base management
packages in the market
 Each organization uses models somewhat differently
 There are many model classes
 Within each class there are different solution approaches
 Some MBMS capabilities require expertise and reasoning
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-81
5.15 Desirable Capabilities of
MBMS
 Control
The DSS user should be provided with a spectrum of control.
The system should support both fully automated and manual
selection of models that seem most useful to the user for an
intended application. This will enable user to proceed at the
problem-solving space that is most comfortable for his or her
experimental familiarity with the task at hand. It should also
be possible for the user to introduce subjective information.
 Flexibility
The DSS user should be able to develop part of the solution
using one approach and then be to switch to another
modeling approach, if this is preferable.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-82
5.15 Desirable Capabilities of
MBMS
 Feedback
The MBMS of the DSS should provided sufficient feedback to enable the
user to be aware of the state of the problem-solving process at any time.
 Interface
The DSS user should feel comfortable with the specific model from the
MBMS that is in use at any given time. The user should not have to
laboriously supply inputs when her or she does not with to do.
 Redundancy reduction
This can be accomplished by use of shared models and associated
elimination of redundant storage.
 Increased consistency
Through the multiple decision makers using the same model and the
associated reduction of inconsistency that may result from use of different
data or different version of a model.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-83
5.15 MBMS Design Must Allow the
DSS User to:
1. Access and retrieve existing models.
2. Exercise and manipulate existing models
Including model instantiation, selection, synthesis, and
provision of suitable inputs
3. Store existing models
Including model representation, abstraction, and physical
and logical storage.
4. Maintain existing models
as appropriate for changing conditions
5. Construct new models with reasonable effort
When they are needed, usually by building new models
using existing models is as building block.
6. Results should be interpreted and analyzed.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-84
5.15 MBMS other respects
 Modeling languages
Number of language used to perform the optimization and
simulation. LINGO (LINDO System), AMPL (Fourer)
GAME (Brooke), etc.
 Relational MBMS
With a relational view of data, a model is viewed as a virtual
file or virtual relation: execution, optimization, sensitivity
analysis.
 Object-oriented model base and its management
Using OO views, OOMBS is possible to build so that
maintain logical independence between the model and the
other DSS components, facilitating intelligent and stabilized
integration of the components.
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-85
5.15 MBMS other respects
 Models for database and MIS design and their
management
Models describing efficient database and MIS design are
using in that the deployed systems will function in the best
way. A model is developed to describe and evaluate a
nonexistent aspect of the business. When the system is
deployed, it functions as if the decision makers have had
many years of experience in running the new system. Thus,
the model building and evaluation are training tools for the
DSS team members.
 Enterprise and Business Process Reengineering Modeling
and Model Management System
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-86
SUMMARY
 Models play a major role in DSS
 Models can be static or dynamic
 Analysis is under assumed certainty, risk, or
uncertainty
Influence diagrams
Spreadsheets
Decision tables and decision trees
 Spreadsheet models and results in influence diagrams
 Optimization: mathematical programming
 Linear programming: economic-based
 Heuristic programming
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-87
SUMMARY
 Simulation - more complex situations
 Expert Choice
 Multidimensional models - OLAP
 Quantitative software packages-OLAP (statistical, etc.)
 Visual interactive modeling (VIM)
 Visual interactive simulation (VIS)
 MBMS are like DBMS
 AI techniques in MBMS
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-88
Assignments (individual)
Access the Microsoft Web site (http://www.microsoft.com/)
and find information about Excel. Also, access Comshare’s
Web site (http://www.comshare.com/) to explore Visual
IFPS/Plus for the PC and Expert Choice Inc.’s Web site
(http://www.expertchoice.com/) to explore Expert choice.
Write a report about characteristics (such as software functions,
using environments, properties, and performance vs. price,
whether needed future programming etc.) of each of three
software. State your recommendations in what
environments to use which software.(no more than 1000
Chinese words)
Review concepts and tools in this chapter:
1)
2)
3)
4)
What are the major modeling issues?
How many category of models? What are their characteristics. Give an
example for each kind of model type.
Decision tree and decision table, simulation, Heuristic programming.
What are the Desirable Capabilities of MBMS?
Fall, 2006
All right reserved, YAO Zhong, School of E&M, BUAA
5-89
Descargar

Chapter 5 Modeling and Analysis