0% found this document useful (0 votes)
22 views33 pages

Classification

Uploaded by

learningshubh26
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
22 views33 pages

Classification

Uploaded by

learningshubh26
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 33

Unit-5

Classification

1
Outline
 Introduction to classification
 Classification & prediction issues
 Classification methods
Introduction to classification
 Classification is a supervised learning method.
 It is a data mining function that assigns items in a collection to
target categories or classes.
 The goal of classification is to accurately predict the target class
for each case in the data.
 For example, a classification model could be used to identify loan
applicants as low, medium, or high credit risks.
 In supervised learning, the learner(computer program) is provided
with two sets of data, training data set and test data set.
 The idea is for the learner to “learn” from a set of labeled
examples in the training set so that it can identify unlabeled
examples in the test set with the highest possible accuracy.
Introduction to classification (Cont..)
 Suppose a Database D is given as D = {t1,t2,..tn} and a set of desired
classes are C={C1,…,Cm}.
 The classification problem is to define the mapping m in such a
way that which tuple of database D belongs to which class of C.
 Actually we divides D into equivalence classes.
 Prediction is similar, but it viewed as having infinite number of
classes.
Classification Example
 Teachers classify students grades as A,B,C,D or E.
 Identify individuals with credit risks (high, low, medium or
unknown).
 In cricket (batsman, bowler, all-rounder)
 Websites (educational, sports, music)
Classification Example (Cont..)
 How teachers give grades to students based on their obtained
marks?
• If x >= 90 then A grade.
• If 80 <= x < 90 then B grade. x
<90 >=90
• If 70 <= x < 80 then C grade.
• If 60 <= x < 70 then D grade. x A
<80 >=80
• If x < 60 then E grade.
x B
<70 >=70
x C
<60 >=60
E D
Classification : a two step process
1) Model Construction
Classification
Algorithms
Training
Data

Name Rank Years Tenured


Classifier
Mike Asst. Prof. 3 No (Model)
Mary Asst. Prof. 7 Yes
Bill Prof. 2 Yes
Jim Asso. Prof. 7 Yes If Rank = ‘professor’
Dave Asst. Prof. 6 No OR year > 6
THEN tenured = ‘yes’
Anne Asso. Prof. 3 No
Classification : a two step process (Cont..)
2) Model Usage
Classifier
Testing
Data

Unseen
Data
Name Rank Years Tenured
Tom Asst. Prof. 2 No
Merlisa Asso. Prof. 7 No
(Jeff, Professor, 4)
George Prof. 5 Yes
Joseph Asst. Prof. 7 Yes
Tenured?

Yes
Classification : a two step process (Cont..)
1) Model Construction
• Describing a set of predetermined classes :
o Each tuple/sample is assumed to belong to a predefined class, as determined by
the class label attribute.
o The set of tuples used for model construction is called as training set.
o The model is represented as classification rules, decision trees, or
mathematical formulae.

2) Model Usage
• For classifying future or unknown objects
o Estimate accuracy of the model
o The known label of test sample is compared with the classified result from the
model.
o Accuracy rate is the percentage of test set samples that are correctly classified
by the model.
Classification & prediction issues
Data Preparation
 Data cleaning
• Pre-process data in order to reduce noise and handle missing values.
 Relevance analysis (Feature selection)
• Remove the irrelevant or redundant attributes.
 Data transformation
• Generalize the data to higher level concepts using concept hierarchies and/or
normalize data which involves scaling the values.
Classification & prediction issues (Cont..)
Evaluating Classification Methods
 Predict accuracy
• This refers the ability of the model to correctly predict the class label of new or
previously unseen data.
 Speed and scalability
• Time to construct model
• Time to use the model
 Robustness
• Handling noise and missing values
 Interpretability
• Understanding and insight provided by model
 Goodness of rules
• Decision tree size
• Strongest rule or not
Classification methods
Classification
Decision Tree

Bayesian
Classification

Rule Based
Classification

Neural Network
Decision tree
 One of the most common tasks is to build models for the
prediction of the class of an object on the basis of its attributes.
 The objects can be seen as a customer, patient, transaction, e-mail
message or even a single character.
 Attributes of patient object can be heart rate, blood pressure,
weight and gender etc.
 The class of the patient object would most commonly be
positive/negative for a certain disease.
Decision tree (Cont..)
 In decision tree are represented by a fixed set of attributes (e.g.
gender) and their values (e.g. male, female) described as
attribute-value pairs.
 If the attribute has small number of disjoint possible values (e.g.
high, medium, low) or there are only two possible classes (e.g.
true, false) then decision tree learning is easy.
 Extension to decision tree algorithm also handles real value
attributes (e.g. salary).
 It gives a class label to each instance of dataset.
Decision tree (Cont..)
 Decision tree is a classifier in the form of a tree structure
• Decision node: Specifies a test on a single attribute
• Leaf node: Indicates the value of the target attribute
• Arc/edge: Split of one attribute
• Path: A disjunction of test to make the final decision
Decision tree representation- example

Root Node

Branches

Leaf Node Leaf Node

Set of Possible Answers Set of Possible Answers


Key requirements for classification
 Sufficient data:
• Enough training cases should be provided to learn the model.

 Attribute-value description:
• Object or case must be expressible in terms of a fixed collection of
properties or attributes (e.g., hot, mild, cold).

 Predefined classes (target values):


• The target function has discrete output values (boolean or multiclass)
Important terms for decision tree
 Entropy
 Information Gain
 Gini Index
Entropy (E)
 It defines the certainty of a decision
• 1 if completely certain,
• 0 if completely uncertain,
• Normally data remains between 0 to 1 as entropy, a probability-based
measure used to calculate the amount of uncertainty.
Entropy (E) (Cont..)
 It measures that of how much information we don't know (how
uncertain we are about the data).
 It can be also used to measure how much information we gain
from an attribute when the target attribute is revealed to us.
 Which attribute is best?
✔ The attribute with the largest expected reduction in entropy is the 'best'
attribute to use next.
✔ Because if we have a large expected reduction it means taking away that
attribute has a big effect, meaning it must be very certain.
Entropy (E) (Cont..)
 A decision tree is built top-down from a root node and involves
partitioning the data into subsets that contain instances with
similar values (homogenous).
 ID3 algorithm uses entropy to calculate the homogeneity of a
sample.
 If the sample is completely homogeneous the entropy is zero and
if the sample is an equally divided it has entropy of one.
Information Gain
 Information gain can be used for continues-valued (numeric)
attributes.
 The attribute which has the highest information gain is selected
for split.
 Assume, that there are two classes P(positive) & N(negative).
 Suppose we have S samples, out of these p samples belongs to
class P and n samples belongs to class N.
 The amount of information, needed to decide split in S belongs to
P or N & that is defined as
-
Gini Index
 Assume there exist several possible split values for each attribute.
 We may need other tools, such as clustering, to get the possible
split values.
 It can be modified for categorical attributes.
 An alternative method to information gain is called the Gini Index.
 Gini is used in CART (Classification and Regression Trees).
 If a dataset T Contains examples from n classes, gini index, gini(T)
is defined as
Gini (T) = 1 - 2
• n: the number of classes
• pj: the probability that a tuple in D belongs to class Ci
Bayesian Classification
 Thomas Bayes, who proposed the Bayes Theorem so, it named
Bayesian theorem.
 It is statistical method & supervised learning method for
classification.
 It can solve problems involving both categorical and continuous
valued attributes.
 Bayesian classification is used to find conditional probabilities.
The Bayes Theorem
 The Bayes Theorem:
• P(H|X)=
 P(H|X) : Probability that the customer will buy a computer given that we know
his age, credit rating and income. (Posterior Probability of H)
 P(H) : Probability that the customer will buy a computer regardless of age,
credit rating, income (Prior Probability of H)
 P(X|H) : Probability that the customer is 35 years old, have fair credit rating and
earns $40,000, given that he has bought computer (Posterior Probability of X)
 P(X) : Probability that a person from our set of customers is 35 years old, have
fair credit rating and earns $40,000. (Prior Probability of X)
Naïve Bayes classifier - Example
Age Income Student Credit_Rating Class : buys_computer
<=30 High No Fair No
<=30 High No Excellent No
31..40 High No Fair Yes
>40 Medium No Fair Yes
>40 Low Yes Fair Yes
>40 Low Yes Excellent No
31..40 Low Yes Excellent Yes
<=30 Medium No Fair No
<=30 Low Yes Fair Yes
>40 Medium Yes Fair Yes
<=30 Medium Yes Excellent Yes
31..40 Medium No Excellent Yes
31..40 High Yes Fair Yes
>40 Medium No Excellent No
Naïve Bayes classifier - Solution
Age
P (<=30 | Yes) = 2/9 P (<=30 | No) = 3/5
P (31..40 | Yes) = 4/9 P (31..40 | No) = 0/5 P (Yes) = 9/14
P (> 40 | Yes) = 3/9 P (> 40 | No) = 2/5 P (No) = 5/14
Income
P (High | Yes) = 2/9 P (High | No) = 2/5
P (Medium | Yes) = 4/9 P (Medium | No) = 2/5
P (Low | Yes) = 3/9 P (Low | No) = 1/5
Student
P (No | Yes) = 3/9 P (No | No) = 4/5
P (Yes | Yes) = 6/9 P (Yes | No) = 1/5
Credit_rating
P (Fair | Yes) = 6/9 P (Fair | No) = 2/5
P (Excellent | Yes) = 3/9 P (Excellent | No) = 3/5
Naïve Bayes classifier - Solution
 An unseen sample Y = (<=30, Low, Yes, Excellent)
 P(Y|Yes).P(Yes) = P(<=30|Yes). P(Low|Yes). P(Yes|Yes).
P(Excellent|Yes) . P(Yes)
= 2/9 * 3/9 * 6/9 * 3/9 * 9/14
= 0.010582

 P(Y|No).P(No) = P(<=30|No). P(Low|No). P(Yes|No).


P(Excellent|No) . P(No)
= 3/5 * 1/5 * 1/5 * 3/5 * 5/14
= 0.005142
 Choose the class so that it maximizes this probability, this means
that new instance will be classified as Yes (Buys_computer)
Try yourself (Bayesian Classification)
Car No Color Type Origin Stolen? Unseen Data
1 Red Sports Domestic Yes Y = <Red, Domestic,
SUV>
2 Red Sports Domestic No Actual Data
3 Red Sports Domestic Yes Y = <Red, Sports,
Domestic>
4 Yellow Sports Domestic No
5 Yellow Sports Imported Yes 0.024,0.072
(Unseen)
6 Yellow SUV Imported No 0.192, 0.096
7 Yellow SUV Imported Yes (Actual)
8 Yellow SUV Domestic No
9 Red SUV Imported No
10 Red Sports Imported Yes
Rule Based Classification
 It is featured by building rules based on an object attributes.
 Rule-based classifier makes use of a set of IF-THEN rules for
classification.
 We can express a rule in the following from
• IF condition THEN conclusion
 Let us consider a rule R1,
R1: IF age=youth AND student=yes THEN buy_computer=yes
• The IF part of the rule is called rule antecedent or precondition.
• The THEN part of the rule is called rule consequent (conclusion).
• The antecedent (IF) part the condition consist of one or more attribute tests
and these tests are logically ANDed.
• The consequent (THEN) part consists of class prediction.
Rule Based Classification (Cont..)
 We can also write rule R1 as follows:
R1: ((age = youth) ^ (student = yes)) => (buys_computer = yes)
• If the condition (that is, all of the attribute tests) in a rule antecedent holds
true for a given tuple, we say that the rule antecedent is satisfied and that
the rule covers the tuple.
 A rule R can be assessed by its coverage and accuracy.
 Given a tuple X, from a class labeled data set D, let it covers the
number of tuples by R; the number of tuples correctly classified by
R; and |D| be the number of tuples in D.
 We can define the coverage and accuracy of R as
Coverage (R) = Accuracy (R) =
Neural Network
 Neural Network is a set of connected INPUT/OUTPUT UNITS, where each
connection has a WEIGHT associated with it.
 Neural Network learning is also called CONNECTIONIST learning due to the
connections between units.
 Neural Network learns by adjusting the weights so It is able to correctly
classify the training data and after testing phase, to classify unknown data.
 Strengths of Neural Network:
• It can handle against complex data. (i.e., problems with many parameters)
• It can handle noise in the training data.
• The Prediction accuracy is generally high.
• Neural Networks are robust, work well even when training examples contain
errors.
• Neural Networks can handle missing data well.

You might also like