SOFT COMPUTING (SWE1015)

Module 5: FUZZY LOGIC AND APPROXIMATE

REASONING

Dr Chiranji Lal Chowdhary

VIT Vellore
 SYLLABUS

• Fuzzy truth values

• Fuzzy propositions
• Fuzzy rules:
• Formation of rules
• Decomposition of rules
• Aggregation of rules
• Fuzzy reasoning
• FIS
• Fuzzy Decision Making
 LOGIC
 It is an analytical theory of art of reasoning
• Goal is to systematize and codify principles of valid reasoning

 Father of Logic: Greek philosopher , Aristotle(384-322 B.C.)

 The modern symbolic logic started with the book “Begriff ss

chrift (1879)”
 by Gottlob Frege (1848 – 1925)(Father of Modern Logic)
 who developed a system of logic for use in his study of the
foundations of arithmetic
 LOGICAL CONNECTIVES
 Consider the proposition p and q whose truth values belong to
the truth value set {0, 1}. The meaning of the logical
connectives is given by definitions and expressed by equations
in which p and q stand for the truth values of the propositions
p and q.

 Negation we have ~ p = 1− p
 Conjunction we have p  q = min( p, q)
 Disjunction we have p  q = max( p, q)
 Implication we have p → q = min(1,1 + q − p)
 Equivalence we have p  q = 1− | p − q |
 SOME IMPORTANT PROPERTIES

• CONTRADICTION: p p

• TAUTOLOGY: p p

• MODUS PONENS: p  ( p → q) → q

• MODUS TOLENS: q  ( p → q) → p
 FUZZY LOGIC
• Founder is L.A.Zadeh (started in 1973)
• Introduced the idea of Linguistic variables and compositional
rules of inference

• Fuzzy logic as the name suggests, is the logic underlying

models of reasoning which are approximate rather than exact

• Fuzzy logic provides an effective conceptual framework for

dealing with the problem of knowledge representation in an
environment of uncertainty and imprecision
 FUZZY LOGIC BASIC (CONTD….)
• A precise definition telling us what fuzzy logic is does not exist

• Although there is not a unique system of knowledge called

fuzzy logic its meaning can be explained

• There is a correspondence (isomorphism) between classical

sets and classical logic

• Fuzzy sets are a generalization of classical sets and infinite-

valued logic is a generalization of classical logic.
• There is a correspondence (isomorphism) between these two
areas.
 FUZZY LOGIC BASICS (CONTD…..)
 Fuzzy logic uses as a major tool fuzzy set theory

 Basic mathematical ideas for fuzzy logic evolve from infinite-

valued logic, thus there is a link between the two logics

 Fuzzy logic can be considered as an extension of infinite-

valued logic in the sense of incorporating fuzzy sets and fuzzy
relations into the system of infinite –valued logic.

 FUZZY LOGIC FOCUSES ON LINGUISTIC VARIABLES IN NATURAL

LANGUAGES AND AIMS TO PROVIDE FOUNDATIONS FOR
APPROXIMATE REASONING WITH IMPRECISE PROPOSITIONS
 Evolvement of Fuzzy Logic

Classical
Infinite-
Logic valued
Logic

Fuzzy
Logic

Fuzzy Sets
Classical
Sets

Fuzzy Numbers
 DOMAIN OF FUZZY LOGIC
• Fuzzy logic deals with:

1.Linguistic variables
2.Linguistic modifiers
3.Propositional fuzzy logic
4.Inferential rules
5.Approximate reasoning
 LINGUISTIC VARIABLES
• Linguistic Variables: Those variables whose values are words
or sentences in natural or artificial languages
• Example: The word ‘Age’ used in natural language
• Age is a linguistic variable taking values very young, young,
middle age, old and very old, called the terms of the linguistic
variable age.
• The terms are fuzzy sets
• Each term is defined by a membership function.
• Consider the universal set U = [0, 100]
MEMBERSHIP FUNCTIONS OF TERMS FOR
LINGUISTIC VARIABLE “OLD”

very young ( x) = 1 for 0  x  5,

= (30 − x) / 25 for 5  x  30.

 young ( x) = ( x − 5) 25 for 5  x  30,

= (50 − x) 2 0 for 30  x  50

middle age ( x) = ( x − 30) 20 for 30  x  50,

= (70 − x) 2 0 for 50  x  70.
 MEMBERSHIP FUNCTIONS OF TERMS FOR
LINGUISTIC VARIABLE “OLD” CONTD…
• Continuing

old ( x) = ( x − 50) 20 for 50  x  70,

= (90 − x) 20 for 70  x  90.

very old ( x) = ( x − 70) 20 for 70  x  90,

=1 for 90  x  100.
 Terms of the linguistic variable age

 AGE

M
e Very Young Young middle age old Very old
m
b
e
r
s
h
i
p
X
0 5 30 50 70 90 100

Age in years→
 LINGUISTIC MODIFIERS
 Let x U and A is a fuzzy set with membership function  A ( x)
 We denote by m a linguistic modifier, for instance very, not
fairly etc.
 Then by mA we mean a modified fuzzy set by m whose
membership function mA ( x) is a composition of a suitable
function f(x) and  A ( x)
expressed as mA ( x) = f (  A ( x)).
 Examples:
 Not: f(x) = 1 – x, notA ( x) = 1 −  A ( x),
 Very: f(x)= ( x)2 veryA ( x) = [  A ( x)]2
 Fairly: f(x)= ( x)1/ 2  fairlyA ( x) = [  A ( x)]1/ 2

 TRUTH (Baldwin, 1979)
• The most important linguistic variable
• It is described by a fuzzy set with membership function
• Truth and its terms have been defined differently in fuzzy
logic. The simplest definition is
true ( x) = x, x  [0,1]

• The modifiers applied to

true = {( x, true ( x)) x  [0,1], true ( x) = x, x  [0,1]}.
not true ( x) =  false ( x) = 1 − x
very true ( x) = [ true ( x)]2 = x 2 .
 fairly true ( x) = [ true ( x)]1/ 2 = x1/ 2 .
 APPROXIMATE REASONING
 Approximate reasoning uses fuzzy sets and fuzzy logic to
model human reasoning.
 It lacks the precision of the exact reasoning in classical logic
but it is more effective dealing with complex and ill-defined
systems.
 It is an active area of research with some topics still under
discussion and debate:
1.Qualification & Quantification rules
2.Compositional rules for inference
3.Truth modifications
 EXAMPLE
 Classical Modus ponens: p  ( p → q) → q
 This is an implication tautology
 In order to assign inference validity to modus ponens we
consider it as a procedure for deriving true propositions
 Such a rule expressed in symbolic form with the scheme called
syllogism
 Premise 1 p
 Premise 2 p→q
 Conclusion q
 EXAMPLE (CONTD….)

• More detail form:

Premise 1: x is A
Premise 2: If x is A then y is B
Conclusion: y is B
• (Say) x = Socrates, A= a man, B = mortal
• Premise 1: Socrates is a man
• Premise 2: If Socrates is a man then Socrates is
mortal
• Conclusion: Socrates is mortal
 TAUTOLOGY

• A statement is said to be a tautology iff it assumes the truth

value TRUE for all possible assignment of truth values to its
primary components
• Modus Ponens is a Tautology
• We have the following tautological equivalences:
F p p
FpF
T  p T
Tp p
 EXAMPLE (CONTD….)

• More detail form:

Premise 1: x is A
Premise 2: If x is A then y is B
Conclusion: y is B
• (Say) x = Socrates, A= a man, B = mortal
• Premise 1: Socrates is a man
• Premise 2: If Socrates is a man then Socrates is
mortal
• Conclusion: Socrates is mortal
 GENERALISED MODUS PONENS

 Premise 1: p’ (p’ is approximately equal to p)

 Premise 2: p→q
 Conclusion: q’ (q’ is approximately equal to q)
Equivalently
 Premise 1: x is A’
 Premise 2: If x is A then y is B
 Conclusion: y is B’
 Example
 Premise 1: This car is slightly more fast
 Premise 2: If a car is fast then the car is dangerous
 Conclusion: This car is little more dangerous
 RULE BASED SYSTEMS
• Rules: Expressions of the type IF Antecedent THEN
Consequent
• Rule Based System: A system which represents knowledge
using such rules

• Newell and Simon developed one of the most popular rule

based systems in 1972
• Book-> Human Problem Solving: Allen Newell, Herbert
Alexander Simon, Prentice-Hall, 1972 - Education - 920 pages

• In which rules were used to model human problem solving

behaviour
 RULE BASED SYSTEMS

• Two major motivations behind the development of rule

based systems:
➢ Creation of programs that act to reproduce human
behaviour
➢ The production of expert systems

• The components of a rule based system:

1. A working memory
2. A rule base
3.An inference engine
 WORKING MEMORY (WM)

• It is the storage medium in a rule based system.

• It comprises of a set of facts known about a domain
• Its contents represent the current state of the system
• Its components are called the working memory elements
(WMEs)
• With the help of WM the rules communicate with each other
• The initial WM contains the starting status of a rule based
system
• The rules are fired when the conditions in the antecedent are
satisfied.
• The firing of a rule changes the status of the rule based
system
 WORKING MEMORY CONTD…

• This may lead to the change in the working memory by

generating new WMEs and may also generate new rules.

• The firing of a rule may

➢ Generate new elements

➢ Modify existing elements

➢ Delete existing elements

 RULE BASE (RB)

• A RB contains several rules of the form given above

• The RB provides the power of reasoning to the system
• The most general structure of a rule may be of the form:
• IF (Condition 1 AND Condition 2 AND… Condition m) THEN
(Action 1 OR Action 2 OR…Action k)
• The reasoning is carried out by firing one or more rules in the
system
• At a time the current state of the working memory may make
several of the rules to be capable of being fired
• A situation when only one such rule can be fired creates no
problem
 CONFLICT RESOLUTION STRATEGIES

• When more than one rule can be fired, it leads to a conflict

among the rules
• Conflict resolution policies are used to solve such a situation
• Some of the conflict resolution strategies are:
• Order
• Specificity
• Recency
• Refractoriness
• In order strategy, the first rule in order of presentation is
fired. This strategy is one of the most common ones. In
PROLOG this strategy is followed.
 SPECIFICITY

• Among the rules whose antecedents are satisfied, select the

one which has the most specific conditions. For example, if
there are rule sets like
• IF (elephant) THEN add colour (grey)
• IF (elephant) AND (royal) THEN add colour (white)
• IF (elephant) AND (African) THEN add ears (large)
• It is clear that the last two rules are more specific than the
first rule. So, these two rules are given preference over the
first rule.
 RECENCY

• Under recency, there are several options

• Sometimes the rule that matches on the most recently

created working memory element is fired

• In the other extreme, the rule which was least recently used
is fired
 REFRACTORINESS

• Under refractoriness, the rule which was fired with the same
set of values of the variables is not selected

• The advantage of this approach is that it helps in avoiding the

looping of reasoning, which occurs as a consequence of firing
a rule repeatedly

• IT MAY BE NOTED THT OFTEN A COMBINATION OF THE

ABOVE PRINCIPLES ARE USED TO RESOLVE CONFLICT
 INFERENCE ENGINE

• The inference engine is a process

• It uses the rules in a rule based system and the information
stored in a working memory to derive new information
about a given problem
• Given the contents of the working memory, the inference
engine determines a set of rules which can be fired
• These are those rules whose antecedents are satisfied under
the present scenario
• As mentioned above, if the rule set contains only one
element then that rule is fired
 INFERENCE ENGINE CONTD…

• If this rule set contains more than one rule, one or more of
the conflict resolution methods are used to determine the
specific rule which can be fired
• If a rule is fired then its consequent is carried out

• This may lead to addition of a working memory element or

modify such a working memory element or add one or more
new rules to the rule base

• After one such firing again matching is carried out to find a

fresh set of rules which can be fired.
• This process is continued until no other rule can be fired
 CONVENTIONAL PROGRAMES vs RULE BASED
SYSTEMS
• The following points describe the differences between
conventional programs and a rule based system

• The first major difference is the declarativeness of the rule

based systems, which is not there in conventional systems

• In a rule based system, facts and rules are declared and also
there is an inference mechanism to control the system

• Both the conventional programs and the rule based systems

use procedural interpretation
 CONVENTIONAL PROGRAMES vs RULE BASED
SYSTEMS
• However, the use of the procedural interpretation in case of
rule based systems differ in the point that they can modify
the working memory and as a result other rules move to the
active state and can be fired

• This does not happen in a conventional program

• In a rule based system; each rule represents a unique piece
of knowledge
• As a result, the addition of a new rule does not affect the
system
• But in conventional programs the addition of new
procedures will affect the existing procedures.
 FUZZY PROPOSITIONS

• We have introduced the concept of fuzzy propositions in

earlier slides

• Types of simple fuzzy propositions:

• Unconditional and unqualified propositions
• Unconditional and qualified propositions
• Conditional and unqualified propositions
• Conditional and qualified propositions
 UNCONDITIONAL AND UNQUALIFIED
PROPOSITIONS
• In case there is no condition and only there is a consequent
in the form of If… then….
• So, only there is a fuzzy proposition F in the consequent
• The general format of such a fuzzy proposition ‘p’ is a
sentence given by ‘ X is F’,
• X is a variable that takes values x from some universal set U
• F is a fuzzy set on U that represents a fuzzy predicate
 EXAMPLE

• Example: 1. X is tall 2.The city X is expensive 3. X is young

• In the first case, U may be a given set of people

• In the second case U may be a set of cities under
consideration
• In case 3, U can again be a set of people under consideration
• In the first case F denotes the fuzzy set representing the
fuzzy predicate ‘tall’
• In the second case F may denote a fuzzy set representing the
fuzzy predicate ‘expensive’
• In the final case F may be representing the fuzzy predicate
‘young’
 EXAMPLE CONTD…

• Given a particular value ‘a’ of X, a membership grade F(a) is

assigned to it by the fuzzy set F and this membership grade is
called the degree of truth T(p) of the fuzzy proposition p = ‘ X
is F’ for X = a
• So, we write T(p) = F(a)
• Example: Height is tall. Here, X: Height and F: tall
• So, the degree of truth T(p) of the proposition depends on
the actual value of the height and on the definition of the
predicate “tall”
• For example, if the height is 150 cms then T(p) = 0.7
• So, the role of function T is to provide us with a bridge
between fuzzy sets and fuzzy propositions
 UNCONDITIONAL AND QUALIFIED
PROPOSITIONS
• In this case also there is no condition like the previous case
• Consequent is a qualified fuzzy proposition
• The general format of such a fuzzy proposition ‘p’ is
• A sentence given by ‘X is F is S’ (The qualification of F is ‘is S’)
• where X and F are as in the previous case
• S is a fuzzy truth qualifier
• The propositions of this type can also be represented as
“pro{X is F} is P”
 UNCONDITIONAL AND QUALIFIED
PROPOSITIONS
• Pro: probability of
• P is a fuzzy probability qualifier
• Both S and P are represented by fuzzy sets on [0, 1]

• Some examples of a truth-qualified proposition are:

• X is tall is fairly true ( F: tall, S: Fairly true)
• The city X is beautiful is very true (F: beautiful, S: very true)
• Hari is talented is fairly false (F: talented, S: fairly false)
 UNCONDITIONAL AND QUALIFIED
PROPOSITIONS
• Example:
• Suppose height of Ram is 165 cms
• He belongs to the set representing the predicate tall with the
membership grade 0.75
• This means, in turn that the degree of truth of our truth-
qualified proposition “Ram is tall is fairly true” is 0.75
• For any particular value ‘a’ of X, the degree of truth T(p) of
any truth-qualified proposition p is given by the equation
• T(p) = S(F(a)) ………. (*)
 UNCONDITIONAL AND QUALIFIED
PROPOSITIONS CONTD…
• If we define a new simple predicate G(a) = S(F(a)) for any ‘a’
• ‘a’ being a truth-qualified proposition of the form as in (*)
• IT CAN BE REPRESENTED AS THE UNQUALIFIED
PROPOSITION “X is G”

• In other words, going in the reverse manner

• Unqualified propositions are special truth-qualified
propositions, in which the truth qualifier S is assumed to be
true
 CONDITIONAL AND UNQUALIFIED
PROPOSITIONS
• In this case there is a condition unlike the previous two cases
and the consequent is a fuzzy proposition F
• The general format of such a fuzzy proposition ‘p’ is a
sentence given by
• Format:
• If X is E then Y is F
• Here X and Y are variables whose values are in the sets U and
V respectively (Domain of X is U and that of Y is V)
• E and F are fuzzy sets on U and V respectively
 CONDITIONAL AND UNQUALIFIED
PROPOSITIONS CONTD…
• These propositions of this form may be viewed as
• Propositions of the form ‘(X, Y) is R’, where R is a fuzzy set on
X Y
• The membership function  R of R is given in terms of the
membership functions  A and  B of A and B respectively
by the formula, R ( x, y) = G(  A ( x),  B ( y))
for each x  X and y  Y
• Here, G denotes a binary operation on [0, 1] representing a
suitable fuzzy implication, say the Lukasiewicz implication
ef132132

CONDITIONAL AND UNQUALIFIED

PROPOSITIONS
• The Lukasiewicz implication is given by
G(e, f) = min (1, 1-e+f).
• Let us define the two fuzzy sets E and F by
• E = {(.5, e1 ), (.8, e2 ), (.3, e3 ), (.1, e4 )} and F = {(.3, f1 ), (.1, f 2 ), (.8, f3 )}
• Then using the formula for R in the previous slide, R is given
by
• {.8 / (e1, f1 ),.6 / (e1, f2 ),1/ (e1, f3 ),.5 / (e2 , f1 ),.7 / (e2 , f2 ),1/ (e2 , f3 ),1/ (e3 , f1 ),.8 / (e3 , f2 ),1/ (e3 , f3 )}
• Fuzzy proposition ‘p’ has the T(p) values given by:
X e1 e1 e1 e2 e2 e2 e3 e3 e3
Y f1 f2 f3 f1 f2 f3 f1 f2 f3
T(p) .8 .6 1 .5 .7 1 1 .8 1
 CONDITIONAL AND UNQUALIFIED
PROPOSITIONS CONTD…
• If we define a new simple predicate G(a) = S(F(a)) for any a , a
truth-qualified proposition can be represented as the
unqualified proposition “X is G”

• In other words, going the reverse manner, unqualified

propositions are special truth-qualified propositions, in
which the truth qualifier S is assumed to be true
 CONDITIONAL AND QUALIFIED PROPOSITIONS
• In this case there is a condition and the consequent is a
qualified fuzzy proposition
• The general format of such a fuzzy proposition ‘p’ is a
sentence given by
• If X is E then Y is F is S
• Here X and Y are variables whose values are in the sets U
and V respectively
• E and F are fuzzy sets on U and V respectively
• S is a fuzzy truth qualifier
• The propositions of this type can also be represented as
“pro{X is E|Y is F} is P”
• Here ‘pro’ stands for ‘probability of’, P is a fuzzy probability
qualifier and Pro {X is A|Y is B} is a conditional probability
 FUZZY RULE BASED SYSTEMS

• There are numerous complex phenomena, which can only be

approximately described
• There are ill-defined phenomena also, which cannot be
analysed through crisp or conventional approaches
• These phenomena can be expressed through variables,
which do not have their values as numbers but words or
sentences in a natural or artificial language
• Examples:
• Tajmahal is beautiful
• Einstein was a highly talented man
• Too much of rain brings in flood in rivers
 FUZZY RULE BASED SYSTEMS

• Highlighted words in the above sentences are values of

linguistic variables
• These words can be represented as fuzzy sets and their
values are labels of the corresponding fuzzy set
• Quantified variables using conventional approaches may
provide an accurate value being assigned
• On the other hand, linguistic variables provide less precise
but more realistic means of system analysis
• Thus the applicability of linguistic variables in societal
systems is much more realistic than the quantified variables
 FUZZY RULE BASED SYSTEMS

• Traditional approaches to analyse social systems were

dependent upon well-structured mathematical models insist
on precision rather than permitting approximation

• In order to handle the situations in social systems and analyze

them we need to sacrifice exactness and more precisely
numerical values by the way permitting linguistic variables

• This may provide approximate characterization of a social

system but will provide a more realistic analysis and
reasoning process through the use of fuzzy logic in which the
truth values are Linguistic instead of being numeric
 FUZZY RULE BASED SYSTEMS

• There are many special linguistic terms like very, more or

less, extremely or fairly, which are used to modify other
linguistic terms. Such terms are called linguistic hedges.
• Hedges are also used to modify fuzzy predicates.
• This in turn helps in modifying fuzzy truth values and also
fuzzy probabilities.
• The fundamental difference between classical propositions
and fuzzy propositions is in the range of their truth values
• While each proposition (crisp) is required to be either true or
false, the truth or falsity of fuzzy propositions is a matter of
degree
 FUZZY RULE BASED SYSTEMS

• Degree of truth of each fuzzy proposition is expressed by a

number in the unit interval [0, 1]
• Fuzzy proposition p: x is F
• can be modified to another fuzzy proposition ‘Hp’
• by introducing a linguistic hedge H in it as ‘x is HF’
• The unary operations that represent linguistic hedges are
called modifiers
• Given a fuzzy predicate F on a domain X and any x  X , the
modified fuzzy predicate HF for a linguistic hedge H is given by
• HF(x) = h(F(x)),
• where h is the unary operation corresponding to H
 FUZZY RULE BASED SYSTEMS

• A linguistic variable is characterised by a quintuple

• (X, T(X), U, G, M), where

• X: The name of the variable

• T(X): The term-set of X, that is, the collection of its linguistic
values
• U: is the universe of discourse
• G: is a syntactic rule for generating the terms in T(X)
• M(X): is a semantic rule which associates with each term in
T(X) its meaning M(X)
 EXAMPLE

• linguistic variable named Height

• Height = tall + very tall + not tall + very very tall + not very tall
+…+short + very short + not short + …+ not very tall and not
very short + … extremely tall + … + more or less tall
• U(Tall) = [0, 200], measured in cm
• A value of height, that is, tall may be viewed as a name of a
fuzzy subset of U which is characterised by its compatibility
function c: U → [0, 1], with c(u) representing the
compatibility of a numerical height u with the label tall
• For example, the compatibilities of the numerical height 190,
150 and 40 with tall might be 1, 0.7 and 0, respectively
 FUZZY RULE BASED SYSTEMS

• In the expression for height in the previous slide, among the

terms we have the primary terms, for example, tall and short
• The meanings of these terms are both subjective and
context-dependent and hence must be defined a priori
• We have used the modifier ‘not’, two sentential connectives
‘and’ and ‘or’ and also linguistic hedges like ‘very’, ‘more’,
‘less’, ‘extremely’ and ‘quite’.
• Some of the standard methods to define the compatibility
functions of linguistic variable X modified by using hedges
‘very’ and ‘more or less’ are presented below.
cveryX (u ) = (c X (u )) 2 cmore or less X (u ) = (c X (u ))1/ 2
 FUZZY RULE BASED SYSTEMS

cnot X (u) = 1 − cX (u)

cXandY (u ) = cX (u )  cY (u )

cXorY (u ) = cX (u )  cY (u )
• If ctall (70) = 0.2 and cshort (70) = 0.1 then
cverytall (70) = 0.04 cvery short (70) = 0.01
c (70) = 0.96 cnot very short (70) = 0.99
not verytall

cnot verytall and not very short (70) = 0.96

 MODIFIERS AND THEIR CHARACTERISTICS

• The types of modifiers are defined as follows:

• For any modifier ‘m’
• ’m’ is said to be strong if m(a) < a, for all a [0,1]
• ‘m’ is said to be weak if m(a) > a, for all a [0,1]
• ‘m’ is called the identity modifier if m(a) = a, for all a [0,1]
• A strong modifier reduces the truth value of the associated
proposition as the predicate to which it is attached gets
stronger
• A weak modifier increases the truth value of the proposition
as the predicate to which it is attached gets weakened
 MODIFIERS AND THEIR CHARACTERISTICS

• Example: Let us consider the three sentences

• P: Ram is tall
• Q: Ram is very tall
• R: Ram is fairly tall
• Here, the linguistic variable ‘tall’ is modified by using two
hedges; ‘very’ and ‘fairly’
• We assume that the two hedges ‘very’ and ‘fairly’ are given by
two modifiers m1 and m2 given by m1 (a) = a 2 and
• m2 (a) = a respectively
• m1 is a strong modifier where as m2 is a weak modifier
 FORMATION OF FUZZY RULES

• The general way of representing human knowledge is by

forming natural language expressions given by
• IF antecedent TEN consequent
• There are tree general forms that exist for any linguistic
variable
• 1. Assignment statements
• 2. Conditional statements
• 3. Unconditional statements
• ASSIGNMENT STATEMENTS:
• These are
• Y = small
 FORMATION OF FUZZY RULES CONTD…

• Orange colour = orange

• a=s
• Climate = autumn
• Outside temperature = normal
• CONDITIONAL STATEMENTS:
• IF y is very cool THEN stop
• IF A is high THEN B is low ELSE B is not low
• IF temperature is high THEN climate is hot
• UNCONDITIONAL STATEMENTS:
• Goto sum
 FORMATION OF FUZZY RULES CONTD…

• Stop
• Turn the pressure low
• Generally both unconditional and conditional statements
place some restrictions on the consequent of the rule based
process
• The restriction statements, irrespective of conditional or
unconditional statements are usually connected by linguistic
connectives such as “and”, “or” or “else”.
• The restrictions denoted as R1, R2, … Rn apply to the
consequent of the rules
 DECOMPOSITION OF COMPOUND RULES

• A compound rule is a collection of many simple rules

combined together
• Any compound rule structure may be decomposed and
reduced to a number of simple canonical rule forms
• The rules are generally based on natural language
representations
• The methods:
• 1. Multiple conjunctive antecedents
• 2. Multiple disjunctive antecedents
• 3.Conditional statements (with ELSE and UNLESS)
• 4. Nested IF-THEN rules
 MULTIPLE CONJUNCTIVE ANTECEDENTS

• IF x is A1, A2, … An THEN y is Bm

• Assume a new fuzzy subset Am is defined as
• 𝐴𝑚 = 𝐴1 ∩ 𝐴2 ∩. . .∩ 𝐴𝑛 and is expressed by their
membership function as
• 𝜇𝐴𝑚 (𝑥) = min[𝜇𝐴1 (𝑥), 𝜇𝐴2 (𝑥), . . . 𝜇𝐴𝑛 (𝑥)൧
• In view of the fuzzy intersection operation, the compound
rule may be rewritten as
• IF Am THEN Bm
 MULTIPLE DISJUNCTIVE ANTECEDENTS

• IF x is A1 OR A2… OR An THEN y is Bm
• This can be written as
• IF x is An then y is Bm
• Where the fuzzy set Am is defined as
• 𝐴𝑚 = 𝐴1 ∪ 𝐴2 ∪. . . 𝐴𝑛 is defined by
• 𝜇𝐴𝑚 (𝑥) = max[𝜇𝐴1 (𝑥), 𝜇𝐴2 (𝑥), . . . 𝜇𝐴𝑛 (𝑥)൧
 CONDITIONAL STATEMENTS (WITH ELSE and
UNLESS)
• Statements of the kind
• I. IF A1 THEN (B1 ELSE B2)
• Can be decomposed into two simple canonical rule forms,
connected by “OR”
• (IF A1 THEN B1) OR (IF NOT A1 THEN B2

• II. IF A! (THEN B!) UNLESS A2

• Can be decomposed as
• (IF A1 THEN B1) OR IF A1 THEN NOT B1
 CONDITIONAL STATEMENTS (WITH ELSE and
UNLESS)
• III. IF A1 THEN B1 ELSE IF A2 THEN B2
• (IF A! THEN B1) OR
• IF NOT A1 AND IF A2 THEN B2
 NESTED_IF_THEN RULES

• The Rule
• IF A1 THEN [ IF A2 THEN B1] can be written in the form
• IF A! AND A2 THEN B1
• So, based upon all the rules above compound rules can be
decomposed into series of canonical simple rules
 AGGREGATION OF FUZZY RULES

• The rule based system involves more than one rule

• Aggregation of rules is the process of obtaining the overall
consequents from the individual consequents provided by
each rule
• The methods adopted are:
• I. Conjunctive system of rules
• II. Disjunctive system of rules
 CONJUNCTIVE SYSTEM OF RULES

• Applicable for a system of rules to be jointly satisfied

• The rules are connected by “and” connectives
• Here the aggregated output ‘y’ is determined by the fuzzy
intersection of all the individual consequents; yi, I = 1, 2, …n
• y = y1 and y2…..and yn
• 𝑦 = 𝑦1 ∩ 𝑦2 . . .∩ 𝑦𝑛
• The aggregated output can be defined by the membership
function
• 𝜇𝑦 (𝑦) = min[𝜇𝑦1 (𝑦), 𝜇𝑦2 (𝑦), . . . 𝜇𝑦𝑛 (𝑦)], ∀𝑦 ∈ 𝑌
 CONJUNCTIVE SYSTEM OF RULES

• Applicable when at least one of the rules is required

• The rules are connected by “or” connectives
• Here the aggregated output ‘y’ is determined by the fuzzy
union of all the individual consequents; yi, I = 1, 2, …n
• y = y1 or y2…..or yn
• 𝑦 = 𝑦1 ∪ 𝑦2 ∪. . . 𝑦𝑛
• The membership function of the aggregate output is given by
• 𝜇𝑦 (𝑦) = max[𝜇𝑦1 (𝑦), 𝜇𝑦2 (𝑦), . . . 𝜇𝑦𝑛 (𝑦)൧
 FUZZY INFERENCE SYSTEM

• A fuzzy inference system uses “IF…THEN” rules along with

connections “OR” or “AND” for making necessary decision
rules
• The inputs may be crisp or fuzzy
• The output is always a fuzzy set
• Only when FIS is used as a controller its outputs should be
crisp
• This necessitates the inclusion of a defuzzification unit for
converting fuzzy variables into crisp variables along FIS
 CONSTRUCTION OF A FIS

• Its functional blocks are:

1. A rule base that contains numerous fuzzy IF-THEN rules
2. A database that defines the membership functions of fuzzy
sets used in fuzzy rules
3. A decision-making unit that performs operation on the rules
4. Fuzzification interface unit that converts the crisp quantities
into fuzzy quantities
5. A defuzzification interface unit that converts the fuzzy
quantities into crisp units
 BLOCK DIAGRAM OF FIS
• g

Knowledge Base
Data Base Rule Base

Input
Output
Fuzzification Defuzzification
Interface Unit Interface Unit
Crisp Crisp

Decision
Fuzzy making Unit Fuzzy
 WORKING METHODOLOGY

• STEP-1: The crisp inputs are converted into a fuzzy input

through a fuzzy interface unit
• STEP-2: Database and rule base are called together as
knowledge base. It is framed
• STEP-3: Defuzzification process is carried out to produce crisp
output
• STEP-4: The fuzzy rules are formed in the rule base and
suitable decisions are made in the decision making unit
 MAMDANI FIS (1975)

• Fuzzy sets are used as rule consequents

• It was first proposed to control a steam engine and a boiler
combination by synthesizing (combine elements of several
sources—to help you make a point) a set of fuzzy rules
obtained from people working on the system
• ALGORITHMIC STEPS:
• STEP-1: Determine a set of fuzzy rules
• STEP-2: Make the inputs fuzzy by using input membership
functions
• STEP-3: Combine the fuzzified inputs according to the fuzzy
rules for establishing a rule strength
 MAMDANI FIS CONTD…

• STEP-4: Determine the consequent of the rule by combining

the rule strength and the output membership function
• STEP-5: Combine all the consequents to get an output
distribution
• STEP-6: Finally, a defuzzified output distribution is obtained
 DECISION MAKING

• Making decision is undoubtedly one of the most fundamental

activities of human beings

• We all are faced in our daily life with varieties of alternative

actions to take

• The beginnings of decision making, as a subject of study, can

be traced, presumably, to the late 18th century, when various
studies were made in France regarding methods of election
and social choice
 DECISION MAKING CONTD…

• Since these initial studies, decision making is based largely on

• Theories and
• Methods developed in this century is enormous

• The subject of decision making is

• The study of how decisions are actually made and
• How they can be made better and more successfully
 DECISION MAKING CONTD…

• Much of the focus in developing the field has been

• In the area of management
• In which the decision-making process is of key importance
• Functions such as
• Inventory control
• Investment
• personnel actions
• new product development
• allocation of resources and
• many others
 DECISION MAKING CONTD…

• Decision making is an integral part of management planning,

organizing, controlling and motivation processes.

• The selection of one strategy over others depends on some

criteria, like utility, sales, cost, return etc.

• The decision should be made whenever the organisation or

an individual faces a problem of decision making or
dissatisfied with the existing decisions or when alternative
decisions are specified.
 DECISION MAKING CONTD…

• Classical decision making generally deals with

• A set of alternative states of nature
• A set of alternative actions that are available to the decision
maker
• A relation indicating the state or outcome to be expected
from each alternative action
• A utility or objective function, which orders the outcomes
according to their desirability
 DECISION MAKING CONTD…

• A decision is said to be made under conditions of certainty

when the outcome for each action can be determined and
ordered precisely

• The alternative that leads to the outcome yielding the

highest utility is chosen
• That is, the decision making problem becomes an
optimization problem, the problem of maximizing the utility
function
 CLASSES OF DECISION MAKING PROBLEMS

• Several classes of decision-making problems are usually

recognized.
• These classifications depend upon their criteria for doing so
• These are:
• Individual decision making and Multiperson decision making
• This classification depends upon the number of decision
makers
• Whether it is only one person or a group of decision makers
involved in the process
 CLASSES OF DECISION MAKING PROBLEMS

• Single criterion decision making and Multicriteria decision

making
• This characterisation depends on
• Whether a simple optimisation of a utility function is done
under constraints or an optimization is done under multiple
objective criteria
• Single stage decision making and multistage decision making
• This classification is decided on whether decision making can
be done at a single stage or can be done iteratively in several
stages
 FUZZY DECISION MAKING

• In the conventional approach to decision making, the principal

ingredients of a decision process are
• A set of alternatives
• A set of constraints on the choice between different
alternatives
• A performance function which associates with each
alternative the gain (or loss) resulting from the choice of that
alternative
 FUZZY DECISION MAKING

• Much of the decision-making in the real world takes place in

an environment in which
• The goals, the constraints and the consequences of possible
actions are not known precisely
• To deal quantitatively with imprecision, we usually employ the
concepts and techniques of probability theory and
information theory
• In doing so, we are tacitly accepting the premise that
imprecision – whatever its nature- can be equated with
randomness
• However, randomness and fuzziness can be differentiated
• By fuzziness we mean a type of imprecision which is
associated with fuzzy sets
 FUZZY DECISION MAKING

• Essentially, randomness has to do with uncertainty

concerning membership or non-membership of an object in
a crisp set
• Whereas fuzziness deals with classes in which there may be
grades of membership of objects in a fuzzy set
• This distinction between the two concepts of randomness and
fuzziness we can say that
• Fuzziness is quite different from probability theory
• The concept of fuzzy decision making was introduced and
studied by Bellman and Zadeh in their seminal paper in 1970
 FUZZY DECISION MAKING

• Decision processes in which fuzziness enters in one way or

another can be studied from many points of view
• This leads to the introduction of the three basic concepts
• Fuzzy goal
• Fuzzy constraint
• Fuzzy decision
• Roughly speaking:
• A fuzzy goal is an objective
• which can be characterised as a fuzzy set in an appropriate
space
 AN EXAMPLE

• Suppose, we have the set of alternatives X be Z, the set of

integers.
• Then a fuzzy goal G may be expressed in terms of its
membership function defined over X
• It can be like
• The alternate x should be much smaller than 100
• The concept of fuzzy goal provides significant advantages to a
fuzzy decision making frame work than the performance
function used in the case of a normal decision process
 FUZZY DECISION MAKING

• A fuzzy decision situation in this model is characterised by

the following components:
• A set of possible actions: A
• A set of goals
• Each goal is expressed in terms of a fuzzy set defined on A
• A set of constraints
• Each of which is also expressed by a fuzzy set defined on A
• Thus, in this process of decision making,
• Relevant goals and constraints are expressed in terms of
fuzzy sets
• A decision is determined by an appropriate aggregation of
these fuzzy sets
 FUZZY BAYESIAN DECISION MAKING

• In classical Bayesian decision making method the future states

of the nature are characterised as probability events
• The sum of the probabilities is equal to 1
• Let the set of possible states of nature be 𝛿 = {𝑠1 , 𝑠2 , . . . 𝑠𝑛 ሽ
• Then the vector representation of the probabilities of these
states is
𝑛
• P = 𝑃(𝑠1 ), 𝑃(𝑠2 ), . . . 𝑃(𝑠𝑛 ) , where ෌𝑖=1 𝑃(𝑠𝑖 ) =1
• These probabilities are called prior probabilities
• The decision maker can choose from “m” alternatives
𝐴 = {𝑎1 , 𝑎2 , . . . 𝑎𝑚 ሽ
 FUZZY BAYESIAN DECISION MAKING CONTD…

• For a given alternative 𝑎𝑗 , a utility value for a future state of

nature 𝑠𝑖 to occur is taken to be 𝜇𝑗𝑖 .
• The decision maker determines these utility values
• These values express the value or cost of each alternative
state pair for each 𝑎𝑗 , 𝑠𝑖 combination
• The expected utility with the jth alternative is given by
𝑛
• 𝐸𝑋(𝑢𝑗 ) = ෍ 𝑢𝑗𝑖 𝑃(𝑠𝑖 ൯
𝑖=1
• The common decision criterion is the maximum expected
utility among all the alternatives 𝐸𝑋(𝑢∗ ) = max𝐸𝑋(𝑢𝑗 ൯
𝑗
• This leads to the selection of alternative 𝑎𝑘 if 𝑢∗ = 𝐸𝑋(𝑢𝑘 )