8 Multiple Criteria Decision

Making (MCDM)

 ZHAN, Wenjie (Professor) 詹文杰

 School of Management, Huazhong Unive
rsity of Science & Technology
 Tel: 027-87556472
 Email: [email protected]
 8 Multiple Criteria Decision
Making (MCDM)
 Objectives:
 To know the Structure of Multiple objectives
 To deal with Incommensurable units
 To deal with Conflict among criteria
 Sensitive analysis to weight
 To elicit weight with Analytic Hierarchy Proc
ess (AHP)
 TOPSIS method
 8 Multiple Criteria Decision
Making (MCDM)
 8.1 Introduction to MCDM
 8.2 Process of MCDM
 8.3 Structure of Objectives
 8.4 How to deal with Incommensurable units
 8.5 How to deal with Conflict among criteria
 8.6 Weighted Sum Model vs. Weighted Produ
ct Model
 8.7 How to elicit weight: AHP
 8.8 TOPSIS method
 8.1 Introduction to MCDM
 Multiple-criteria decision-making (MCDM) or
multiple-criteria decision analysis (MCDA) is a sub-
discipline of operations research that explicitly
evaluates multiple conflicting criteria in
decision making (both in daily life and in settings
such as business, government and medicine).
 MCDM encompasses both MADM (multi-attribute
decision making) and MODM (multi-objective
decision making).
 MCDA relates to a number of discrete alternatives.
 MODM relates to an infinite or uncountable number of
alternatives.
 8.1 Introduction to MCDM
 Conflicting criteria are typical in evaluating
options: cost or price is usually one of the main
criteria, and some measure of quality is typically
another criterion, easily in conflict with the cost.
 Example 1: In purchasing a car, cost, comfort, safety,
and fuel economy may be some of the main criteria we
consider – it is unusual that the cheapest car is the most
comfortable and the safest one.
 Example 2: In portfolio management, we are interested
in getting high returns but at the same time reducing our
risks, but the stocks that have the potential of bringing
high returns typically also carry high risks of losing money.
 8.1 Introduction to MCDM
 The core MCDA is useful for:
 Dividing the decision into smaller, more understandable
parts;
 Analyzing each part;
 Integrating the parts to produce a meaningful solution.

 MCDA problems are comprised of five components:

 1. Goal
2. Decision maker with opinions (preferences)
3. Decision alternatives
4. Evaluation criteria (interests)
5. Outcomes or consequences associated with
alternative/interest combination
 8.1 Introduction to MCDM:
Example 3
Car Price ( ￥ 000s) Fuel Eff. Comfort Index
(mpg)
Mercedes (M) 400 25 10
Chevrolet (C) 150 28 3
Toyota (T) 250 35 6
Volvo (V) 350 30 8
It is easy to make decision with single objective, such as
price, fuel efficiency, comfort Index.
But in the real world, people always make their decision with
multiple objectives, which means that they take into account
the price, fuel efficiency, comfort Index at the same time.
 8.1 Introduction to MCDM:
Example 3
 Two things make multiple objectives decision making
difficult:
 (1) Conflict among criteria.
 (2) Incommensurable units: Each objective/attribute has a
different unit of measurement.

Car Price ( ￥ 000s) Fuel Eff. Comfort Index

(mpg)
Mercedes (M) 400 25 10
Chevrolet (C) 150 28 3
Toyota (T) 250 35 6
Volvo (V) 350 30 8
 8.2 Process of MCDM:
an Example (1/8)
MCDM process: Car buying example
Identifying
a Problem
I need to buy
a new car.

Identifying •Price
Decision Criteria •Interior Comfort
•Durability
•Repair Record
•Performance
Allocating Weights •Price 10
to Criteria •Interior Comfort 8
•Durability 6
•Repair Record 4
•Performance 2
 Decision-making process: Car buying example (2/8)
Developing
Alternatives Dodge Audi Isuzu Chevy

Toyota Ford Jeep Mazda

Analyzing Toyota •Price Dodge

Alternatives •Comfort
Ford Audi
•Durability
Jeep •Repair Record Isuzu
Mazda •Performance Chevy

Selecting
an Alternative
The Toyota
is the best.

Implementing
the Choice Appraising
Decision Results
Decision-making process: Some comments (3/8)

1) Problem Identification
 Solving the wrong problem perfectly is no better than
do nothing for the right problem
 Problem is identified by comparing the current state
with some standards that represent a desired state of
affaires:
 Past performance, previously set goals, performance of other
units/organizations …
 Problem identification is subjective in nature
Decision-making process: Some comments (4/8)

2) Decision Criteria
 They reflect the factors that decision
makers think important in making the
choice
 They are not equally important
 A simple approach to deal with different
criteria is to assign them different weights
 Judgments are involved in selecting
criteria and assigning weighs
 Decision-making process: Some comments (5/8)
3) Developing and Analyzing
Alternatives
INITIAL
INITIAL INTERIOR
INTERIOR DURA-
DURA- REPAIR
REPAIR PERFOR-
PERFOR- HAND-
HAND-
PRICE
PRICE COMFORT BILITY RECORD MANCE LING
COMFORT BILITY RECORD MANCE LING TOTAL
TOTAL
ALTERNATIVES
ALTERNATIVES (10)
(10) (8)
(8) (5)
(5) (5)
(5) (3)
(3) (1)
(1)

Jeep
JeepCherokee
Cherokee 22 20
20 10
10 80
80 88 40
40 77 35
35 55 15
15 55 55 195
195
Ford
FordTaurus
Taurus 99 90
90 66 4848 55 25
25 66 30
30 88 24
24 66 66 223
223
Mercedes
MercedesC230
C230 88 80
80 55 4040 66 30
30 66 30
30 44 12
12 66 66 198
198
Saab
Saab900
900 99 90
90 55 4040 66 30
30 77 35
35 66 18
18 55 55 218
218
Mazda
Mazda626
626 55 50
50 66 4848 99 45
45 10
10 5050 77 21
21 77 77 221
221
Dodge
DodgeIntrepid
Intrepid 10
10100
100 55 40
40 66 30
30 44 20
20 33 99 33 33 202
202
Ford
FordExplorer
Explorer 44 40
40 88 6464 77 35
35 66 30
30 88 24
24 99 99 202
202
Isuzu
IsuzuRodeo
Rodeo 77 70
70 66 4848 88 40
40 66 30
30 55 15
15 66 66 209
209
Volvo
Volvo850`
850` 99 90
90 77 5656 44 20
20 44 20
20 44 12
12 55 55 203
203
Audi 90
Audi 90 55 50 8
50 8 6464 5 25
5 25 4 20
4 20 10 30 10 10 199
10 30 10 10 199
Toyota
ToyotaCamry
Camry 66 60
60 5 40 10 50 10 50
5 40 10 50 10 50 66 18
18 66 66 224
224
Volkswagen
VolkswagenPassat
Passat 88 80 6
80 6 4848 6 30
6 30 5 25
5 25 77 21
21 8 8 212
8 8 212
Decision-making process: Some comments (6/8)
4) Quantitative Tools to
Decision Analysis
 Operations research models
 Linear programming
 Queuing theory
 Economic order quantity (EOQ)
 Financial techniques
 Ratio analysis
 Break-even analysis
 Decision analysis tools
 Decision tree
 Payoff matrix
Decision-making process: Some comments (7/8)

5) Implementation and Evaluation

 Selecting the “best” choice based on the analysis
 Putting into the decision into action includes
 Communicating with those to be affected and
 Gaining their commitment of support
 Participation of decision making brings enthusiasms
 Evaluation of the decision result is an important phase
of decision making process and part of the controlling
function
Decision-making process: Summary (8/8)
Identification of
a problem

Identification of
Decision criteria
Evaluation of decision
Allocation of effectiveness
Weights to criteria

Development of
alternatives

Analysis of
alternatives

The Decision- Selection of an

Making Process alternative

Implementation of
The alternative
 8.3 Structure of Objectives
Overall Objective

Objective 1 Objective 2 Objective n

Sub-objective 1 Sub-objective 1 Sub-objective 1 Sub-objective 1

Attribute for Sub- Attribute for Sub- Attribute for Sub- Attribute for Sub-
objective 1 objective 1 objective n objective 1
 1) Model for MCDM

Figure 1: A Typical Decision Matrix

 2) Basic terms: Objectives
 Objectives :
 An objective has been defined by Keeney a
nd Raiffa as an indication of the preferred
direction of movement.
 Thus, when stating objectives, we use term
s like ‘minimize’ or ‘maximize’.
 Typical objectives might be to minimize cos
ts or maximize market share.
 2) Basic terms: Attributes
 Attributes
 An attribute is used to measure performance in relation
to an objective.
 For example, if we have the objective ‘maximize the
exposure of a television advertisement’ we may use
the attribute ‘number of people surveyed who recall
seeing the advertisement’ in order to measure the
degree to which the objective was achieved.
 Sometimes we may have to use an attribute which is
not directly related to the objective. Such an attribute
is referred to as a proxy attribute. For example, a
company may use the proxy attribute ‘staff turnover’ to
measure how well they are achieving their objective of
maximizing job satisfaction for their staff.
 3) Typical hierarchical
structure of objectives
Overall Objective

Objective 1 Objective 2 Objective n

Sub-objective 1 Sub-objective 1 Sub-objective 1 Sub-objective 1

Attribute for Sub- Attribute for Sub- Attribute for Sub- Attribute for Sub-
objective 1 objective 1 objective n objective 1
 3) First level objective
 The highest level of this structure general
ly represents the broad overall objectives
that are instrumental in initiating the mul
tiple objective decision problem in the firs
t place.
 These objectives are, however, often vag
uely stated and, hence, unoperational.
 3) Second level objectives
 As we go down the hierarchical level,
objectives at the lower level are more specific
and more operational than those in the higher
level.
 They are perceived as means to achieving
higher ends represented by objectives in the
higher level.
 Thus objectives at the lowest level of the
hierarchy are “most specific” and “most
operational”.
 4) Attributes for objectives
 An objective is operational is there is a
practical way to assess the level of achieving
such an objective.
 To facilitate this practical method, a set of
attributes is assigned to each objective in the
lowest level.
 An attribute is a measurable quantity whose
value reflects the degree of achievement for a
particular objective.
 4) Hierarchical structure of
objectives for the choosing a Car
Maximum satisfaction car to buy

Minimum price Maximum fuel eff. Maximum comfort

400, 150, 250, 350 25, 28, 35, 30 10, 3, 6, 8

Car Price ( ￥ 000s) Fuel Eff. (mpg) Comfort Index

Mercedes (M) 400 25 10

Chevrolet (C) 150 28 3

Toyota (T) 250 35 6
Volvo (V) 350 30 8
 5) Example: Choosing a Location
for a Municipal Aquatics Facility (1/7)
 Goal: Choose a site for a 40,000 sq. ft. municipal
aquatics facility in a city of 120,000 people.
 Decision Makers: 10-member Parks and Recreation
Advisory Board.
 Decision Alternatives: Decide among four alternative
sites:
 Hawes Tract – a vacant site on the rapidly growing outskirts
of town
 Chatham Street – an infill site near the center of town
 Bartley Park – a new park on an old agricultural site in the
south
 North Cary Park – a site adjacent to a community park and a
well-traveled riparian greenway
 5) Example: Choosing a Location
for a Municipal Aquatics Facility
(2/7)
 Identify Stakeholder Interests ：
 Board member interests serve as the criteria by which to evaluate
each site.  Interests are identified through a facilitated discussion
of the board members.
 The Parks and Recreation Advisory Board members identified 10
key interests, many of which were further broken down in
to sub-interests.
 Interests and Sub-Interests
Identified by Board Members (3/7)
 1 ） Proximity to other facilities
 a. Maintain distance from year-round aquatics
 b. Maintain distance from seasonal aquatic centers
 c. Maintain distance from community centers
 d. Be accessible to public parks
 e. Be accessible to high schools
 2 ） Proximity to users
 a. Be easily accessible to existing neighborhoods
 b. Be easily accessible to planned neighborhoods to be built in the
near-term
 3 ） Ease of access
 a. From surrounding neighborhoods
 b. By greenway
 d. By roadway
 Interests and Sub-Interests
Identified by Board Members (4/7)
 4 ） Environmental impact 
 a. Minimize disruption to mature trees and vegetation
 b. Avoid topographic challenges
 c. Minimize impact to watershed
 5 ） Traffic Impact
 a. New traffic can be accommodated by existing street network
 6 ） Compatibility with Surrounding Area
 a. Facility footprint and use must be compatible with surrounding
neighborhood
 7 ） Visibility and Profile of Facility
 a. Facility footprint must be compatible with, and not detract from
surrounding neighborhood
 Interests and Sub-Interests
Identified by Board Members (5/7)
 8 ） Ability to Meet Future Demand
 a. Tract must be able to accommodate expansion of fa
cility and uses
 9 ） Parking
 a. Tract must be of an appropriate dimension to accom
modate parking
 10 ） Cost
 a. Reduce on-site development costs
 b. Reduce off-site development costs
5) Example: Structure of
objectives (6/7)
 5) Example: Decision Matrix
(7/7)
1 ） Proximity to other facili 2 ） Proximit …. 10 ） Cost
ties y to users

a. b. c. d. e. a. b. …. a. b.
Hawes
Tract 
•1 ） Proximity to other facilit …. 10 ） Cos
ies
t
Chath a. Maintain distance from ye
am ar-round aquatics …. a. Red
uce o
Street  b. Maintain distance from se
n-site
asonal aquatic centers
Bartle
y Park c. Maintain distance from co …. develo
pment
mmunity centers
North costs
Cary
d. Be accessible to public par …. b. Red
ks
Park uce of
e. Be accessible to high scho
f-site
ols
 8.4 How to deal with
Incommensurable units
 To solve the problem of incommensurable
units , we need to attribute scales for each ob
jective.
 There are many ways to attribute scales for e
ach objective.
 Nominal Scales
 Ordinal Scales
 Interval Scales
 Ratio Scales
 Proportional Scoring
 Proportional Scoring is the most application
method to attribute scales for each objective.
 Called proportional because scales linearly.

U i (x)  x Worst
Best Worst
 Proportional Scoring for price
Car Price After
( ￥ 000 Prop Price: Best = 150, Worst =400
s)
Mercedes 400 0 UP(C) = 1;
(M) UP(M) = 0;
Chevrolet 150 1 UP(T) = 0.6;
(C) UP(V) = 0.2;
Toyota 250 0.6
(T) Toyota is 0.4 away from best to worst.
Volvo (V) 350 0.2  Volvo is 0.8 away from best to worst.
 Proportional Scoring for fuel
efficiency
Car Fuel After
Eff. Prop Price: Best = 35, Worst =25
(mpg)
Mercedes 25 0 Uf(T) = 1;
(M)
Uf(M) = 0;
Chevrolet 28 0.3 Uf(C) = 0.3;
(C)
Uf(V) = 0.5;
Toyota 35 1
(T) Chevrolet is 0.3 away from best to worst
Volvo (V) 30 0.5 .
 Volvo is halfway between best/worst.
 Proportional Scoring for
comfort Index
Car Comfort After
Index Prop Price: Best = 10, Worst =3

Mercedes 10 1 Uc(M) = 1;
(M)
Uc(C) = 0;
Chevrolet 3 0
Uc(T) = 3/7;
(C)
Uc(V) = 6/7;
Toyota 6 3/7
(T)
Toyota is 4/7 away from best to worst.
Volvo (V) 8 6/7  Volvo is 1/7 away from best to worst
 After the Proportional Scoring

Car Price Fuel Eff. Comfort Index

Mercedes (M) 0 0 1

Chevrolet (C) 1 0.3 0

Toyota (T) 0.6 1 3/7

Volvo (V) 0.2 0.5 6/7

 8.4 How to deal with
Incommensurable units
 Questions:
 Q1:What is assumption of risk attitude to P
roportional Scoring?
 Q2: Is there any relationship between the
method to deal with incommensurable
units and the risk attitude?
 8.5 How to deal with Conflict
among criteria
 Most decisions require tradeoffs among
objectives because they are conflicting.
 To solve the conflict among criteria, we
use weight for each objective.
 Weights may reflect how important the
attribute/objective is to the decision ma
ker.
 1) Example: Weights for
choosing a car
 Suppose:
 Weight for price =0.4
 Weight for fuel efficiency =0.4
 Weight for comfort index =0.2
 Our weights should respect the “Range of the
attribute scales”.
 It means that the buyer thinks that the price
and the fuel efficiency are equal important,
and both of them are more important that the
comfort.
 aggregating values for
choosing a car
Car Price Fuel Comfort Index aggregating
(0.4) Eff. (0.2) value
(0.4)
Mercedes 0 0 1 0.2
(M)
Chevrolet 1 0.3 0 0.52
(C)
Toyota (T) 0.6 1 3/7 0.686
(0.43)
Volvo (V) 0.2 0.5 6/7 0.309
(0.86)
 Weights for choosing a car
 Suppose:
 Weight for price =0.8
 Weight for fuel efficiency =0.1
 Weight for comfort index =0.1
 It means that the buyer thinks that the
price is the most important thing to buy
a car, and the fuel efficiency and the
comfort are equal and less important.
 aggregating values for
choosing a car
Car Price Fuel Comfort Index aggregating
(0.2) Eff. (0.6) value
(0.2)
Mercedes 0 0 1 0.6
(M)
Chevrolet 1 0.3 0 0.26
(C)
Toyota (T) 0.6 1 3/7 0.58
(0.43)
Volvo (V) 0.2 0.5 6/7 0.66
(0.86)
 8.5 How to deal with Conflict
among criteria
 Questions:
 Q1: Does the weight for each objective
affect the result of decision?
 Q2:The weight for each objective is
whether the subjective or objective
parameter?
 2) Sensitive analysis to weight
 At which range the weight of comfort can
change the result?
 Suppose: the price and the fuel efficiency
are equal important.
 We have:
 wp+wf+wc=1, then wp+wf=1-wc
 wp=wf , then wp=wf=(1-wc )/2
 aggregating values for
choosing a car
Car Price Fuel Comfort aggregating value
(1-wc )/2 Eff. Index
(1-wc )/2 wc
Mercedes 0 0 1 wc
(M)
Chevrolet 1 0.3 0 0.65-0.65*wc
(C)
Toyota (T) 0.6 1 3/7 0.8-0.37*wc
(0.43)
Volvo (V) 0.2 0.5 6/7 0.35+0.51*wc
(0.86)
Sensitive analysis to weight of
comfort

U(T
)

U(C
)
U(V
)

U(M
)
 8.6 Weighted Sum Model vs.
Weighted Product Model
 8.6.1 Weighted Sum Model (WSM)
 8.6.2 Weighted Product Model (WPM)
 8.6.1 Weighted Sum Model
(WSM)
 The weighted sum model (or WSM) is probably
the most commonly used approach, especially in
single dimensional problems.
 If there are M alternatives and N criteria then,
the best alternative is the one that satisfies (in
the maximization case) the following expression:
WSM: an Example
Weighted score/sum method
 Let Sij score of option i using criterion j
 wj weight for criterion j
 Si score of option i is given as:

Sij =  wj Sij

 The option with the best score is selected.

Weighted score/sum method
 The method can be modified by using U
(Sij) and then calculating the weighted u
tility score.
 To use utility the condition of separabili
ty must hold.
 Explain the meaning of separability:
U(Si) =  wj U(Sij)
U(Si)  U( wj Sij)
 The main stages in Weighted
score/sum method
 Stage 1: Identify the decision maker (or decision makers)
. In our problem we will assume that this is just the buye
r of the car, not the seller of the car.
 Stage 2: Identify the alternative courses of action. In our
problem these are, represented by the different brands o
f cars.
 Stage 3: Identify the objectives which are relevant to the
decision problem. In our problem these are the price, the
fuel efficiency, and the comfort index.
 Stage 4: For each objective, assign values to measure th
e performance of the alternatives on that attribute. Then,
deal with Incommensurable units.
 The main stages in Weighted
score/sum method
 Stage 5: Determine a weight for each objective.
This may reflect how important the
objective/attribute is to the decision maker
 Stage 6: For each alternative, take a weighted
average of the values assigned to that
alternative.
 Stage 7: Make a provisional decision.
 Stage 8: Perform sensitivity analysis to see how
robust the decision is to changes in the figures
supplied by the decision maker.
 8.6.2 Weighted Product Model
(WPM)
 The weighted product model (or WPM) is very similar to
the WSM. The main difference is that instead of addition in
the model there is multiplication. Each alternative is
compared with the others by multiplying a number of
ratios, one for each criterion. Each ratio is raised to the
power equivalent to the relative weight of the
corresponding criterion.
 In general, in order to compare the alternatives AK and AL,
the following product has to be calculated:
WPM: an Example
 WPM: An alternative approach
 An alternative approach is one to use only p
roducts without ratios. That is, to use the fo
llowing variants of formula ：
N wj
P( AK )   (a Kj )
j 1
N
P( AK )   w j  (a Kj )
j 1
 8.7 How to elicit weight
 The analytic hierarchy process (AHP) offers a method to
elicit weight when a decision maker is faced with a probl
em involving multiple objectives.
 The method, which was developed by Thomas Saaty wh
en he was acting as an adviser to the US government, ha
s been very widely applied to decision problems in areas
such as economics and planning, energy policy, material
handling and purchasing, project selection, microcomput
er selection, budget allocations and forecasting.
 AHP involves a relatively complex mathematical procedur
e.
 An overview of the AHP
 Stage 1: Make pairwise comparisons of
objectives.
 Stage 2: Transform the comparisons int
o weights by mathematical procedure.
 Stage 3: Check the consistency of the d
ecision maker’s comparisons.
 What is pair-wise comparison?
 Suppose we have two fruits Apple and Banana. I
would like to ask you, which fruit you like better than
the other and how much you like it in comparison
with the other.
 Let us make a relative scale to measure how much
you like the fruit on the left (Apple) compared to the
fruit on the right (Banana).
 What is pair-wise comparison?
 Now suppose you have three choices of fruits. Then
the pair wise comparison goes as the following ：
What is pair-wise comparison?
Since we have 3 objects (Apple, Banana and
Cheery), we have 3 comparisons. Table below
shows the number of comparisons.
Table 7: Number of comparisons
 The Analytic Hierarchy Process
(AHP)
Given values (1-9) Explanation
1 Equal Both objectives have equal importance.

3 Moderate One of the objectives is slightly more

important than the other.
5 Strong One of the objectives is judged as
strongly more important than the other
by experts.
7 Very Strong One of the objectives is judged as very
strongly important compared to the
other.
9 Extreme Importance One alternative is strictly superior to the
other one.
2, 4, 6, 8 Intermediate values Used for compromised judgments when
necessary.

•The scaling is not necessary 1 to 9 but for qualitative data such as preference,
ranking and subjective opinions, it is suggested to use scale 1 to 9.
 Example 1: Choosing a car
(1/7)

 Eliciting the weight in choosing a car.

 Three objectives:
 Price
 Fuel efficiency
 Comfort
Stage 1: Make pairwise compa
risons of objectives. (2/7)

Price Fuel Comfort

Price 1 2 9

Fuel 1/2 1 7

Comfort 1/9 1/7 1

Stage 2: Transform the comparisons into
weights by mathematical procedure (3/7)
 (1) All the quantities in each row are continued
multiplied n times first. Then w* is equal to
the nth power of sum.

wi *  n ai1  ai 2  ...  ain

 (2) Then w* is normalized.

wi *
wi 
w1 *  w2 * ...  wn *
 Stage 2: (4/7)

Price Fuel Comf

ort

Price 1 2 9 w1*  3 1  2  9  2.62

1
Fuel 1/2 1 7 w2 *  3  1  7  1.52
2
Comf 1/9 1/7 1 1 1
ort w3 *  3   1  0.25
9 7
 Stage 2: (5/7)

 Sum of w*=4.39
 Then:
 w1=w1* /4.39=2.62/4.39=0.6
 w2=w2* /4.39=1.52/4.39=0.35
 w3=w3* /4.39=0.25/4.39=0.05
Stage 3: Check the consistency of
the decision maker’s comparisons
(6/7)

 (1) To get the sum of all the quantities in each column.

S j  a1 j  a1 j  ...  anj
 (2) To get check parameter λmax.

max  w1  S1  w2  S 2  ...  wn  S n
 (3) Compare λmax with numbers in the table. If λmax is less than the
corresponding number in the table, it passes the check of consistency.
 Stage 3: (7/7)

Price Fuel Comfort

Price 1 2 9 W1=0.6

Fuel 1/2 1 7 W2=0.35

Comfort 1/9 1/7 1 W3=0.05

S1=1+1/2+1/9 S2=2+1+1/7 S3=9+7+1

=1.61 =3.14 =17

λmax =0.61.61+0.35 3.14+0.05 17= 2.9150 < 3.116

Example 2: School Selection
(1/4)
Pairwise Comparisons among
Objectives (2/4)
Pairwise Comparison of Schools
with each Objective (3/4)
Composition and Synthesis
(4/4)
 8.8 TOPSIS METHOD
 Technique of Order Preference by Similarity t
o Ideal Solution
 In this method two artificial alternatives are h
ypothesized:
 Ideal alternative: the one which has the best level
for all attributes considered.
 Negative ideal alternative: the one which has the
worst attribute values.

 TOPSIS selects the alternative that is the clos

est to the ideal solution and farthest from ne
gative ideal alternative.
Example: ideal solution and
negative ideal alternative
f2 x4 x*

x5
x1
x3
x6

x2
x0

f1
图 9.5 理想解和负理想解示意图
 Input to TOPSIS
 TOPSIS assumes that we have m alternatives (option
s) and n attributes/criteria and we have the score of
each option with respect to each criterion.

 Let xij score of option i with respect to criterion j

We have a matrix X = (xij) mn matrix.
 Let J be the set of benefit attributes or criteria (mor
e is better)
 Let J' be the set of negative attributes or criteria (les
s is better)
Example 1: TOPSIS - Car
buying (1/17)
Weight 0.1 0.4 0.3 0.2

Style Reliability Fuel Eco. Cost

Civic 7 9 9 8

Saturn 8 7 8 7

Ford 9 6 8 9

Mazda 6 7 8 6
 Applying TOPSIS to
Example (2/17)
 m = 4 alternatives (car models)
 n = 4 attributes/criteria

 xij = score of option i with respect to criterion

j
X = {xij} 44 score matrix.
 J = set of benefit attributes: style, reliability,
fuel economy (more is better)
 J' = set of negative attributes: cost (less is be
tter)
 Step 1 of TOPSIS (3/17)
 Step 1(a): calculate (x2ij )1/2 for each
column Style Rel. Fuel
Cost
Civic 49 81 81 64

Saturn 64 49 64 49

Ford 81 36 64 81

Mazda 36 49 64 36
xij2i 230 215 273 230

(x2)1/2 15.17 14.66 16.52 15.17

 Step 1 of TOPSIS (4/17)
 Step 1 (b): divide each column by (x2ij
)1/2 to get rij
Style Rel. Fuel Cost
Civic 0.46 0.61 0.54 0.53

Saturn 0.53 0.48 0.48 0.46

Ford 0.59 0.41 0.48 0.59

Mazda 0.40 0.48 0.48 0.40

 Step 2 of TOPSIS (5/17)
 Step 2 (b): multiply each column by wj
to get vij.
Style Rel. Fuel Cost
Civic 0.046 0.244 0.162 0.106

Saturn 0.053 0.192 0.144 0.092

Ford 0.059 0.164 0.144 0.118

Mazda 0.040 0.192 0.144 0.080

 Step3 of TOPSIS (6/17)
 Step 3 (a): determine ideal solution A*.
A* = {0.059, 0.244, 0.162, 0.080}
Style Rel. Fuel Cost
Civic 0.046 0.244 0.162 0.106

Saturn 0.053 0.192 0.144 0.092

Ford 0.059 0.164 0.144 0.118

Mazda 0.040 0.192 0.144 0.080

 Step 3 of TOPSIS (7/17)

 Step 3 (b): find negative ideal solution

A'.
A' = {0.040, 0.164, 0.144, 0.118}
Style Rel. Fuel Cost
Civic 0.046 0.244 0.162 0.106

Saturn 0.053 0.192 0.144 0.092

Ford 0.059 0.164 0.144 0.118

Mazda 0.040 0.192 0.144 0.080

 Step 4 of TOPSIS (8/17)

 Step 4 (a): determine separation from i

deal solution A* = {0.059, 0.244, 0.162
, 0.080} Si* = [  (vj*– vij)2 ] ½ for ea
ch row j
Style Rel. Fuel Cost
Civic (.046-.059)2 (.244-.244)2 (0)2 (.026)2

Saturn (.053-.059)2 (.192-.244)2 (-.018)2 (.012)2

Ford (.053-.059)2 (.164-.244)2 (-.018)2 (.038)2

Mazda (.053-.059)2 (.192-.244)2 (-.018)2 (.0)2
 Step 4 of TOPSIS (9/17)
 Step 4 (a): determine separation from
ideal solution Si*

(vj*–vij)2 Si* = [  (vj*– vij)2 ] ½

Civic 0.000845 0.029

Saturn 0.003208 0.057

Ford 0.008186 0.090

Mazda 0.003389 0.058

 Step 4 of TOPSIS (10/17)
 Step 4 (b): find separation from negative i
deal solution A' = {0.040, 0.164, 0.144, 0
.118}
Si' = [  (vj'– vij)2 ] ½ for each row
Style j Rel. Fuel Cost
Civic (.046-.040)2 (.244-.164)2 (.018)2 (-.012)2

Saturn (.053-.040)2 (.192-.164)2 (0)2 (-.026)2

Ford (.053-.040)2 (.164-.164)2 (0)2 (0)2

Mazda (.053-.040)2 (.192-.164)2 (0)2 (-.038)2
 Step 4 of TOPSIS (11/17)

 Step 4 (b): determine separation from

negative ideal solution Si'

(vj'–vij)2 Si' = [  (vj'– vij)2 ] ½

Civic 0.006904 0.083

Saturn 0.001629 0.040

Ford 0.000361 0.019

Mazda 0.002228 0.047

 Step 5 of TOPSIS (12/17)

 Step 5: Calculate the relative closeness

to the ideal solution Ci* = S'i / (Si* +S'i )

S'i /(Si*+S'i) Ci*

Civic 0.083/0.112 0.74  BEST

Saturn 0.040/0.097 0.41

Ford 0.019/0.109 0.17

Mazda 0.047/0.105 0.45

Summary: Steps of TOPSIS
(13/17)
 Step 1: Construct normalized decision matrix.

 This step transforms various attribute dimensio

ns into non-dimensional attributes, which allow
s comparisons across criteria.
 Normalize scores or data as follows:

rij = xij/ (x2ij) for i = 1, …, m; j = 1, …, n

i
Summary: Steps of TOPSIS
(14/17)
 Step 2: Construct the weighted normalized de
cision matrix.

 Assume we have a set of weights for each crit

eria wj for j = 1,…n.
 Multiply each column of the normalized decisi
on matrix by its associated weight.
 An element of the new matrix is:

vij = wj rij
Summary: Steps of TOPSIS
(15/17)
 Step 3: Determine the ideal and negative ideal solutio
ns.
 Ideal solution.
A* = { v1* , …, vn*}, where
vj* ={ max (vij) if j  J ; min (vij) if j  J' }
i i

 Negative ideal solution.

A' = { v1' , …, vn' }, where
v' = { min (vij) if j  J ; max (vij) if j  J' }
i i
Summary: Steps of TOPSIS
(16/17)
 Step 4: Calculate the separation measures fo
r each alternative.
 The separation from the ideal alternative is:
Si * = [  (vj*– vij)2 ] ½ i = 1, …, m
j

 Similarly, the separation from the negative id

eal alternative is:
S'i = [  (vj' – vij)2 ] ½ i = 1, …, m
j
Summary: Steps of TOPSIS
(17/17)
 Step 5: Calculate the relative closeness to the
ideal solution Ci*

Ci* = S'i / (Si* +S'i ) , 0  C i*  1

Select the option with Ci* closest to 1.

WHY ?
 In-class Example: recruiting
an assistant (1/7)
 A company wants to recruit a new principal assistant for its in
ternational market. Four candidates have been shortlisted:
Anna, Tom, Jack and Emma.
 Four criteria have been selected to make the decision. As t
he post requires intensive contact with various customers, it i
s necessary for the principal assistant to have strong interper
sonal skills, with the ability to interact effectively with diverse
client styles within different working environments. The role i
nvolves dealing with the international market, and as a result
, extensive experience of living abroad would be advantageo
us. Similar work experience would be beneficial. Each candid
ate is required to sit a written exam to assess their knowledg
e of international culture.
 The performances of each candidate against the four criteria
are shown in Table 8.1.
 In-class Example: recruiting
an assistant (2/7)
Table 8.1 Weights of the criteria and performances of the alternatives.

Interpersonal skills Living abroad Written test Work experience

(score out of 10) (years) (score out of 10) (years)
Weight 0.1 0.4 0.3 0.2
Anna 7 9 9 8
Tom 8 7 8 7
Jack 9 6 7 12
Emma 6 11 8 6
 Step 1: Construct normalized
decision matrix (3/7)
Table 1: Normalized decision matrix
Interpersonal skills Living abroad Written test Work experience
(score out of 10) (years) (score out of 10) (years)
Anna 0.46 0.53 0.56 0.47
Tom 0.53 0.41 0.50 0.41
Jack 0.59 0.35 0.44 0.70
Emma 0.40 0.65 0.50 0.35
 Step 2: Construct the weighted
normalized decision matrix (4/7)
Table 2: Weighted normalized decision matrix
Interpersonal skills Living abroad Written test Work experience
(score out of 10) (years) (score out of 10) (years)
Weight 0.1 0.4 0.3 0.2
Anna 0.046 0.213 0.168 0.093
Tom 0.053 0.165 0.149 0.082
Jack 0.059 0.142 0.131 0.140
Emma 0.040 0.260 0.149 0.070
 Step 3: Determine the ideal and
negative ideal solutions (5/7)
Table 3: Ideal and negative ideal solutions
Interpersonal skills Living abroad Written test Work experience
(score out of 10) (years) (score out of 10) (years)
Anna 0.046 0.213 0.168 0.093
Tom 0.053 0.165 0.149 0.082
Jack 0.059 0.142 0.131 0.140
Emma 0.040 0.260 0.149 0.070
A* 0.059 0.260 0.168 0.140
A’ 0.040 0.142 0.131 0.070
 Step 4: Calculate the separation
measures for each alternative (6/7)
Table 4: Distance calculation
d* d’

Anna 0.068 0.084

Tom 0.113 0.035
Jack 0.124 0.073
Emma 0.075 0.120
Step 5: Calculate the relative clos
eness to the ideal solution Ci*
(7/7)
Table 5: Table of scores
d* d’ C*

Anna 0.068 0.084 0.550

Tom 0.113 0.035 0.240
Jack 0.124 0.073 0.370
Emma 0.075 0.120 0.610
 Homework 7-1:
 The organizers of an international conference
have to choose a hotel that will be the confer
ence venue. The organizers has five choices:
the Alton, the Buttermere, the Castle, the Dor
set and the Elm.
 A value tree was used to identify the attribute
s relating to the decision. These are
 (i) the cost of using the hotel
 (ii) ease of transport to the hotel
 (iii) quality of the conference facilities
 Homework 7-1 (cont.):
•The costs of using the hotels for the conference are given
below. The lower the cost, the better the performance.
•Together with scores that the organizers assigned to them for
the non-monetary attributes , such as transport, facilities. (0 =
worst performance on the attribute, 100 = the best).
 Homework 7-1 (cont.):
 Q1: If the organizers assigned the weights to cost,
transport, facility as 0.2,0.4,0.4. What is the target
hotel?
 Q2: If the organizers compare the importance of cost,
transport, facility as follows. What is the target hotel?

cost transport facility

cost 1 2 9
transport 1/2 1 7
facility 1/9 1/7 1
 Homework 7-2:
 A company is going to choose one of the supplier
among the three according to three criteria: service,
quality, and price. The weights for each of them are
0.2 for service, 0.4 for quality, 0.4 for price. The
normalized decision table is given as follows. Help the
company to choose the supplier according to TOPSIS.
Service Quality Price
(the more, the (the more, the (the less, the better)
better) better)

S1 0.1577 0.7661 0.7840

S2 0.3109 0.6131 0.3666
S3 0.9373 0.1926 0.5009
The end