UNIT -4

By
Dr. Pallavi Malavath
 Syllabus
• Fuzzy Logic: Basic concepts, Fuzzy set theory, Basic
operations, Fuzzification and Defuzzification, Fuzzy
decision making Particle Swarm Optimization,
Genetic Algorithms
• Rough Set Theory: Rough Sets, Differences between
Rough sets and fuzzy sets, Decision Tables, Rule
Induction, Discernibility Matrix
 Crisp vs. Fuzzy Logic

• Crisp Logic (Classical Logic): Values are strictly

true (1) or false (0).
– Example: "The water is hot" → Either True (hot) or
False (not hot).
• Fuzzy Logic: Values range between 0 and 1,
representing degrees of truth.
– Example: "The water is warm" → It could be 0.6
hot and 0.4 cold.
• Membership Functions (MF)
• A Membership Function assigns a degree of belonging (0
to 1) to each input in a fuzzy set.
• Fuzzy Rules (IF-THEN Rules)
• Fuzzy logic systems use IF-THEN rules to make decisions.
• Example: Temperature Control System
• Rule 1: IF temperature is cold, THEN turn heater on high.
• Rule 2: IF temperature is warm, THEN turn heater on
low.
• Rule 3: IF temperature is hot, THEN turn heater off.
Fuzzy set theory
 Basic Fuzzy operations
• Fuzzy operations are performed on fuzzy sets,
whereas crisp operations are performed
on crisp sets. Fuzzy operations are very useful
in the design of a Fuzzy Logic Controller.
• The three basic operations are complement,
intersection, union.
 Fuzzification and Defuzzification.
• Fuzzy logic is widely used in systems where decisions
are made under uncertainty. The two key processes
in fuzzy logic systems are Fuzzification and
Defuzzification.
1. Fuzzification (Crisp to Fuzzy Conversion)
• Fuzzification is the process of converting a crisp
(precise) input into a fuzzy value. This allows
systems to handle imprecise and uncertain data.
• Steps in Fuzzification:
• Identify input variables – Determine the crisp
numerical values that need to be fuzzified.
• Define fuzzy sets (membership functions) –
Represent different linguistic terms such as "Low,"
"Medium," and "High" using membership functions.
• Assign membership values – Each crisp input value is
mapped to a fuzzy set based on its degree of
belonging.
• Example of Fuzzification:
• Imagine a temperature control system where:
• Crisp Input: Temperature = 35°C
• Fuzzy Sets:
– "Cold" (0 to 25°C)
– "Warm" (20 to 40°C)
– "Hot" (30 to 50°C)
• Using a triangular membership function, 35°C belongs:
• 0.25 to "Warm“, 0.75 to "Hot“
 Defuzzification (Fuzzy to Crisp Conversion)
• Defuzzification is the process of converting fuzzy outputs into a
single crisp value, which can be used as a control action in real-
world applications.
• Methods of Defuzzification:
• Centroid Method (Center of Gravity –COG).
• Max Membership Principle (Mean of Maximum - MOM)- takes the
value with the highest membership.
• Weighted Average Method- Averages the fuzzy values weighted by
their membership.
• Mean max membership.
• Center of Sums
• Center of largest area.
• First(or last) of maxima
• [Link] membership principle:
• This is also known as height method. In this
case the defuzzified value is obtained as the
element having the highest membership value
in the domain
• [Link] method: This procedure (also
called of area or the center of gravity) is the
mostly used method of all the defuzzification
methods.
• [Link] average method:
• This is most widely used and computationally
efficient defuzzification methods, it is only
applicable to the symmetrical output
membership functions.
• It takes the weighted arithmetic mean of the
elements in the domain, where the weights are
the membership values of the elements .
• [Link] max membership:
• This is the generalized version of the first
method. Instead of a single point, the
maximum membership may occur at a set of
points forming a plateau in the graph of the
output function.
• This method is also called middle-of-maxima
• [Link] of sums:
• This method is faster and it is not restricted to
symmetric membership functions..
• The computation of the centroids of the
individual membership functions is required in
this process given in the following equation.
6. Center of largest area:
• If the output fuzzy set has at least two convex
sub-regions, obtain the defuzzified value “a”
of the output .
• [Link](or last) of maxima:
In this method the domain Ci with the
maximized membership degree among the
output fuzzy sets Ck, k=1,2…….n, is
determined by using the overall output, which
is nothing but the union of all the individual
fuzzy sets
 Rough Set Theory
(for more info refer 1st unit also)
• Rough Set Theory is a method for dealing with
uncertain data by defining sets with clear and possible
members.
• Rough Set Theory (RST) is a mathematical approach
used to deal with incompleteness in data analysis. It
was introduced by Zdzisław Pawlak in the early 1980s.
• The main advantage of rough set theory in data
analysis is that it does not need any preliminary or
additional information about data − like probability in
statistics.
The proposed approach
•provides efficient algorithms for finding hidden
patterns in data,
•finds minimal sets of data (data reduction),
•evaluates significance of data,
•generates sets of decision rules from data,
•it is easy to understand,
•offers straightforward interpretation of obtained
results,
•most algorithms based on the rough set theory are
particularly suited for parallel processing.
Rough sets consists of:
• Information System: A dataset represented as a table
where rows correspond to objects and columns to
attributes.
• Indiscernibility Relation: Objects with the same values
for a subset of attributes are considered indiscernible.
• Lower and Upper Approximation:
– Lower Approximation: The set of objects that definitely
belong to a concept.
– Upper Approximation: The set of objects that possibly
belong to a concept.
• Boundary Region: The difference between upper and
lower approximations, representing uncertainty.
• Reducts: The minimal subset of attributes preserving
the classification capability of the original dataset.
 Differences between Rough sets and
fuzzy sets
• Rough sets and fuzzy sets are both mathematical tools
used for dealing with uncertainty and imprecision in
data, but they differ fundamentally in their approaches
and applications.
• Example:
• Rough Set Approach: Suppose we are classifying emails
as spam or not spam. If some emails have unclear
classification, they fall into the boundary region of the
rough set.
• Fuzzy Set Approach: In a fuzzy system, an email could be
60% spam and 40% not spam, based on membership
values.
Data Requirements Works well with Requires a predefined
categorical data and membership function
does not require for each fuzzy set.
additional information
like a membership
function.
Application Areas Used in machine Used in control systems,
learning, feature decision-making, and
selection, and approximate reasoning
knowledge discovery where gradual
where uncertainty arises transitions are needed.
due to incomplete
information.
 Differences Between Rough Sets and
Fuzzy Sets
FEATURE ROUGH SETS FUZZY SETS
Based on approximation Based on degrees of
Concept of sets using lower and membership in a set.
upper bounds.
Handling Uncertainty Deals with vagueness by Deals with vagueness by
using boundary regions allowing partial
(lower and upper membership of
approximations). elements.
Membership Membership is crisp Membership is gradual,
(either belongs or does ranging from 0 to 1.
not belong to lower or
upper approximation).
 Decision Table in Rough Set Theory

• A decision table is a tabular representation of

rules or conditions used for decision-making.
In soft computing, decision tables are often
used in rough set theory to extract patterns,
make classifications, and reduce data
dimensionality.
 Components of a Decision Table:

• Condition Attributes – Features or criteria

that influence decision-making.
• Decision Attributes – The outcome or
classification based on condition attributes.
• Objects (Rows) – Instances or cases described
by condition attributes.
• Decision Rules – If-then rules derived from
the table.
 Decision Table in Rough Set Theory:
• Lower Approximation: Objects that definitely belong
to a decision class.
• Upper Approximation: Objects that possibly belong
to a decision class.
• Boundary Region: Objects that cannot be precisely
classified.
• By analyzing inconsistencies in the decision table,
rough set theory helps identify essential attributes
and remove redundant ones.
 Rule Induction
• Rule induction is a process of extracting decision
rules from datasets, commonly used in machine
learning, data mining, expert systems, and rough
set theory.
• It helps in knowledge discovery by identifying
patterns and relationships in data.
 Types of Rule Induction:
• Direct Rule Induction
– Rules are derived directly from the dataset by
identifying patterns.
– Used in decision trees, rough set theory, and
association rule mining.
• Indirect Rule Induction
– Involves learning an intermediate model (neural
networks) and extracting rules from it.
– Used in fuzzy logic systems, neural networks, and
genetic algorithms.
• Rule Induction in Rough Set Theory:
• In rough sets, rule induction is performed
using decision tables where:
• Condition attributes are used to describe
objects.
• Decision attributes determine the outcome.
• Reducts help remove redundant attributes
while preserving classification power.
 Discernibility Matrix
• A discernibility matrix is a tool used in rough
set theory to identify important attributes in a
dataset.
• It helps determine which attributes distinguish
objects from each other, allowing for attribute
reduction while maintaining classification
ability.
• Construction of a Discernibility Matrix
• Given:
• A decision system S = (U, A ∪ {d}), where:
– U is the universe of objects (data instances)
– A is the set of conditional attributes
– d is the decision attribute (class label)
• Each entry M(i, j) contains a set of attributes
that differentiate object x_i from x_j.
• If two objects have the same decision value
(d(x_i) = d(x_ j)), then M(i, j) = ∅ (they are
indiscernible).
• The discernibility matrix is used to construct
reducts, which are minimal subsets of
attributes that preserve classification ability.
• Applications
• Feature Selection: Helps find minimal
attribute subsets that preserve classification
ability.
• Data Preprocessing: Identifies redundant or
irrelevant attributes.
• Rough Set Theory: Used for computing
reducts and core attributes in knowledge
reduction.