Lect5 Classification and Prediction

(1)

Classification and Prediction

Lecture 5/DMBI/IKI83403T/MTI/UI

Yudho Giri Sucahyo, Ph.D, CISA (yudho@cs.ui.ac.id)

Faculty of Computer Science, University of Indonesia

Objectives

`

Introduction

`

What is Classification?

`

Classification vs Prediction

`

Classification vs Prediction

`

Supervised and Unsupervised Learning

`

D t P

ti

`

Data Preparation

`

Classification Accuracy

`

ID3 Algorithm

`

Information Gain

`

Bayesian Classification

`

Predictive Modelling

University of Indonesia

`

Predictive Modelling

2

Introduction

` Databases are rich with hidden information that can be used for making intelligent business decisions.

` Classification and prediction can be used to extract models

d ibi i d l di f d

describing important data classes or to predict future data trends.

` Classification predicts categorical labels Ex: categorize bank

` Classification predicts categorical labels. Ex: categorize bank loan applications Æsafe or risky.

` Prediction models continuous-valued functions. Ex: predict the expenditures of potential customers on computer equipment given their income and occupation.

` Typical Applications:

` Credit approval, target marketing,

` Medical diagnosis, treatment effectiveness analysisg , y

What is Classification? A two step process

What is Classification? – A two-step process

`

Model construction:

` Each tuple is assumed to belong to a predefined class, as determined by one of the attributes, called the class label.

` Data tuples are also referred to as samples, examples, or objects.

` All tuples used for construction is called training set.

` Since the class label of each training sample is providedÆ

supervised learning. In clustering (unsupervised learning),

th l l b l f h t i i l i t k d th the class labels of each training sample is not known, and the number or set of classes to be learned may not be known in advance.

advance.

` The model is represented in the following forms:

(2)

What is Classification? A two step process (2)

What is Classification? – A two-step process (2)

`

The model is used for classifying future or

unknown objects.

` First, the predictive accuracy of the model is estimated

` The known label of test sample is compared with the classified result from the model.

` Accuracy rate is the percentage of test set samples that are correctly ` Accuracy rate is the percentage of test set samples that are correctly

classified by the model.

` Test set is independent of training set otherwise over-fitting (it may have incorporated some particular anomalies of the training data that are not present in the overall sample population) will occur.

` If the accuracy of the model is considered acceptable the

` If the accuracy of the model is considered acceptable, the model can be used to classify future objects for which the class label is not known (unknown, previously unseen data).

( p y )

5

Classification Process (1)

Classification

Training Data

Classification Algorithms

Data

NAM E RANK YEA RS TENURED

M ike Assistant Prof 3 no

Classifier (Model)

M ary Assistant Prof 7 yes

Bill Professor 2 yes

Jim Associate Prof 7 yes Jim Associate Prof 7 yes Dave Assistant Prof 6 no Anne Associate Prof 3 no

IF rank = ‘professor’ OR years > 6

THEN tenured = ‘yes’

University of Indonesia 6

THEN tenured yes

Classification Process (2)

Classifier

Testing

Data Unseen Data

Data

(Jeff, Professor, 4)

NAM E RANK YEARS TENURED

Tom Assistant Prof 2 no

Tenured?

M erlisa Associate Prof 7 no

George Professor 5 yes

Joseph Assistant Prof 7 es

7

Joseph Assistant Prof 7 yes

What is Prediction?

`

Prediction is similar to classification

` First, construct model.

` Second, use model to predict future or unknown objectsp j

` Major method for prediction is regression:

Linear and multiple regression

Non-liner regression

`

Prediction is different from classification

` Classification refers to predict categorical class label.

` Prediction refers to predict continuous value.

(3)

Classification vs Prediction

`

Sending out promotional literature to every new

customer in the database can be quite costly. A more

cos-efficient method would be to target only those new

h

lik l

h

Æ

customers who are likely to purchase a new computer

Æ

classification

.

P d

h

b

f

h

`

Predict the number of major purchases that a customer

will make during a fiscal year

Æ

prediction

.

Supervised vs Unsupervised Learning

`

Supervised learning (classification)

` Supervision: The training data (observations, measurements, etc.) are accompanied by labels indicating the class of the ) p y g observations

` Based on the training set to classify new datag y

`

Unsupervised learning

(clustering)

` We are given a set of measurements observations etc with

` We are given a set of measurements, observations, etc with the aim of establishing the existence of classes or clusters in the data

the data

` No training data, or the “training data” are not accompanied by class labels

by class labels

10

Issues Data Preparation

Issues – Data Preparation

`

Data preprocessing can be used to help improve the

accuracy, efficiency, and scalability of the classification or

prediction process.

`

Data Cleaning

` Remove/reduce noise and the treatment of missing values

`

Relevance Analysis

` Many of the attributes in the data may be irrelevant to the classification or prediction task. Ex: data recording the day of the week on which a bank loan application was filed is unlikely to be relevant to the success of the application

to be relevant to the success of the application.

` Other attributes may be redundant.

` This step is known as feature selection

` This step is known as feature selection.

Issues Data Preparation

Issues – Data Preparation

`

Data Transformation

` Data can be generalized to higher-level concepts.

` Useful fot continuous-valued attributes.

` Income can be generalized Ælow, medium, high.

` Street Æcity.

` Generalization compresses the original training data, fewer input/output operations may be involved during learning.

` Wh i l t k ( th th d i l i

(4)

Comparing Classification Method

` Predictive accuracy

` Speed and scalability

` time to construct the model

` time to use the model

` Robustness

` handling noise and missing values

` Scalability

` efficiency in large databases (not memory resident data)

` Interpretability:

` the level of understanding and insight provided by the model

` Goodness of rules

` decision tree size

` the compactness of classification rules

Classification Accuracy: Estimating Error Rates

` Partition: Training-and-testing

` use two independent data sets, e.g., training set (2/3), test set(1/3)

` used for data set with large number of samples

` Cross-validation

` divide the data set into ksubsamples

` use k-1subsamples as training data and one sub-sample as test p g p data ---k-fold cross-validation

` for data set with moderate size

` Bootstrapping (leave-one-out)

` for small size data

What is a decision tree?

`

A decision tree is a flow-chart-like tree structure.

` Internal nodedenotes a test on an attribute

` Branch represents an outcome of the test

` All tuples in branch have the same value for the tested attribute.

` Leaf node represents class label or class label distribution. `

To classify an unknown sample, the attribute values of the

sample are tested against the decision tree. A path is traced

from the root to a leaf node that holds the class prediction

f h

l

for that sample.

`

Decision trees can easily be converted to classification rules.

15

Training Dataset

`

An Example

Outlook Tempreature Humidity Windy Class_sunny _hot _high _false _N

`

An Example

from Quinlan’s

ID3

sunny hot high false N

sunny hot high true N

overcast hot highg false P

rain mild high false P

rain cool normal false P

rain cool normal tr e N

rain cool normal true N

overcast cool normal true P sunnyy mild highg false N sunny cool normal false P

rain mild normal false P

ild l t P

sunny mild normal true P

overcast mild high true P overcast hot normal false P

16

(5)

A Sample Decision Tree

Outlook Outlook

overcast

sunny _overcast _rain

humidity P windy

high normal true false

N P N P

Decision Tree Classification Methods

Decision-Tree Classification Methods

`

The basic top-down decision tree generation approach

usually consists of two phases:

`

Tree construction

` At start, all the training examples are at the root.

` Partition examples recursively based on selected

` Partition examples recursively based on selected attributes.

`

Tree pruning

`

Tree pruning

` Aiming at removing tree branches that may lead to errors h l if i t t d t (t i i d t t i i when classifying test data (training data may contain noise, outliers, …)

ID3 Algorithm

` All attributes are categorical

` Create a node N;

` if samples are all of the same class C, then ` return N as a leaf node labeled with C

` return N as a leaf node labeled with C

` if attribute-list is empty then

` return N as a leaf node labeled with the most common class

` select split-attribute with highest information gain

` label N with the split-attribute

` f h l A f lit tt ib t b h f N d N

` for each value Aiof split-attribute, grow a branch from Node N

` let Sibe the branch in which all tuples have the value Aifor split- attribute

` if Siis empty then

attach a leaf labeled with the most common class

Else recursively run the algorithm at Node S_i

` until all branches reach leaf nodes

Choosing Split Attribute –

Information Gain (ID3/C4 5) (1)

Information Gain (ID3/C4.5) (1)

`

Assume all attributes to be categorical (discrete-values).

Continuous-valued attributes must be discretized.

`

Used to select the test attribute at each node in the tree.

`

Also called

measure of the goodness of split

.

(6)

Information Gain (ID3/C4 5) (2)

Information Gain (ID3/C4.5) (2)

` Assume that there are two classes,Pand N.

L h f l S l f l P d

` Let the set of examples Scontain pelements of class Pand n elements of class N.

` The amount of information needed to decide if an arbitrary

` The amount of information, needed to decide if an arbitrary example in S belong to Por Nis defined as

I( ) p lo g p n lo g n

` Assume that using attribute A as the root in the tree will partition S {S S S} needed to classify objects in all subtrees S_i :

needed to classify objects in all subtrees S_i :

E A i iI

Information Gain (ID3/C4 5) (3)

Information Gain (ID3/C4.5) (3)

` The attribute Ais selected such that the information gain

gain(A) = I(p, n) - E(A)

is maximal, that is, E(A), , ( )is minimal since I(p, n)(p, )is the same to all attributes at a node.

` In the given sample data attribute outlook is chosen to split at

` In the given sample data, attribute outlook is chosen to split at the root :

i ( tl k) = 0 246 gain(outlook) = 0.246

gain(temperature) = 0.029

gain(humidity) = 0.151

gain(windy) = 0.048

Information Gain (ID3/C4 5) (3)

Information Gain (ID3/C4.5) (3)

`

Examples:

` See Table 7.1.

` Class label: buys_computer. Two values: YES, NO.

` m = 2. C₁ correspond to yes, C₂correspond to no.

` 9 samples of class yes and 5 samples of class no.

` Compute the expected information needed to classify a given sample

Information Gain (ID3/C4 5) (4)

Information Gain (ID3/C4.5) (4)

` Next, compute the entropy of each attribute. Let’s start with the ib

` Using equation (7 2) the expected information needed to classify

` Using equation (7.2), the expected information needed to classify a given sample if the samples are partitioned by age is

694

` Hence, the gain in information from such a partitioning: Gain(age) = I(s₁, s₂) –E (age) = 0.246

14 14

14

( g ) ( ₁ ₂) ( g )

` Similarly, we can compute Gain(income) = 0.029, Gain(student) = 0.151, Gain(Credit_rating) = 0.048.

(7)

How to use a tree?

`

Directly

` test the attribute value of unknown sample against the tree.

` A path is traced from root to a leaf which holds the label

`

Indirectly

` decision tree is converted to classification rules

` one rule is created for each path from the root to a leaf

` IF-THEN is easier for humans to understand

` Example:

IF age = “<=30” AND student = “no” THEN buys_computer = “no”

Tree Pruning

` A decision tree constructed using the training data may have / f

too many branches/leaf nodes.

` Caused by noise, overfitting

M l f l

` May result poor accuracy for unseen samples

` Prune the tree: merge a subtree into a leaf node.

U i f d diff f h i i d

` Using a set of data different from the training data.

` At a tree node, if the accuracy without splitting is higher than the accuracy with splitting replace the subtree with a leaf node the accuracy with splitting, replace the subtree with a leaf node, label it using the majority class.

` Pruning Criterion:

` Pessimistic pruning: C4.5

` MDL: SLIQ and SPRINT

University of Indonesia ` Cost complexity pruning: CART

26

Classification and Databases

`

Classification is a classical problem extensively studied by

` statisticians

` AI, especially machine learning researchers

`

Database researchers re-examined the problem in the

context of large databases

` most previous studies used small size data, and most algorithms are memory resident

`

Recent data mining research contributes to

` Scalability

` Generalization-based classification

` Parallel and distributed processing

Classifying Large Dataset

`

Decision trees seem to be a good choice

` relatively faster learning speed than other classification methods

` can be converted into simple and easy to understand classification rules

` can be used to generate SQL queries for accessing databases

` has comparable classification accuracy with other methods

`

Classifying data sets with millions of examples and a few

`

Classifying data-sets with millions of examples and a few

hundred even thousands attributes with reasonable

speed

(8)

Scalable Decision Tree Methods

`

Most algorithms assume data can fit in memory.

`

Data mining research contributes to the scalability issue,

especially for decision trees.

p

y

`

Successful examples

` SLIQ(EDBT’96 Mehta et al ’96)

` SLIQ(EDBT 96 -- Mehta et al. 96)

` SPRINT(VLDB96 -- J. Shafer et al.’96)

` PUBLIC(VLDB98 Rastogi & Shim’98)

` PUBLIC(VLDB98 -- Rastogi & Shim 98)

` RainForest (VLDB98 -- Gehrke, et al.’98)

Previous Efforts on Scalability

`

Incremental tree construction (Quinlan’86)

(

)

` using partial data to build a tree.

` testing other examples and those mis-classified ones are used g p to rebuild the tree interactively.

`

Data reduction (Cattlet’91)

` reducing data size by sampling and discretization.

` still a main memory algorithm.

`

Data partition and merge (Chan and Stolfo’91)

` partitioning data and building trees for each partition.

` merging multiple trees into a combined tree.

` experiment results indicated reduced classification accuracy.

Presentation of Classification Rules

31

Other Classification Methods

`

Bayesian Classification

`

Bayesian Classification

`

Neural Networks

`

Genetic Algorithm

`

Rough Set Approach

`

Rough Set Approach

`

k-Nearest Neighbor Classifier

`

Case-Based Reasoning (CBR)

`

Fuzzy Logic

`

Fuzzy Logic

`

Support Vector Machine (SVM)

(9)

Bayesian Classification

`

Bayesian classifiers are statistical classifiers.

`

They can predict class membership probabilities, such as

the probability that a given sample belongs to a particular

class.

`

Bayesian classification is based on Bayes theorem.

`

Naive Bayesian Classifier is comparable in performance

with decision tree and neural network classifiers.

`

Bayesian classifiers also have high accuracy and speed

when applied to large databases.

Bayes Theorem (1)

`

Let

X

be a data sample whose class label is unknown.

`

Let

H

be some hypothesis, such as that the data sample

X

belongs to a specified class

C

.

`

We want to determine

P

(

H

|X), the probability the the

hypothesis

H

holds given the observed data sample

X

.

`

P

(

H

|X) is the

posterior probability or a posteriori

probability

, of

H

conditioned on

X

.

` Support the world of data samples consists of fruits, described by their color and shape., Suppose that X is red and round, and

h H i h h h i h X i l Th P(H|X) that H is the hypothesis that X is an apple. Then P(H|X) reflects our confidence that X is an apple given that we have seen that X is red and round.

University of Indonesia seen that X is red and round.

34

Bayes Theorem (2)

`

P

(

H

) is the

prior probability or a priori probability

, of

H.

` The probability that any given data sample is an apple,

regardless of how the data sample looks.

`

The posterior probability is based on more information

(such as background knowledge) than the prior

b bili hi h

i

i d

d

f

X

probability which is independent of

X.

`

Bayes theorem is

)

`

See example 7.4 for example on Naive Bayesian

) (X P

Classification.

Predictive Modeling in Databases

`

What if we would like to predict a continuous value,

rather than a categorical label?

` Prediction of continuous values can be modeled by statistical

h i f i

techniques of regression.

` Example:

` A m dle t redict the salar f c lle e rad ates ith 10 ears f ` A modle to predict the salary of college graduates with 10 years of

work experience.

` Potential sales of a new product given its price.

`

Many problems can be solved by

linear regression

.

`

Software packages for solving regression problems:

`

Software packages for solving regression problems:

(10)

Linear Regression

`

Data are modeled using a straight line.

`

The simplest form of regression

`

Bivariate liner regressions models a random variable

g

Y

(called a

response variable

), as a linear function of another

random variable,

X

(called a

predictor variable

)

` Y = α+ βX

`

See Example 7.6 for an example of linear regression.

p

g

`

Other regression models

` Multiple regression

` Log-linear models

Prediction: Numerical Data

Prediction: Categorical Data

39

Conclusion

` Classification is an extensively studied problem (mainly in

statistics, machine learning & neural networks)

` Classification is probably one of the most widely used data

mining techniques with a lot of applications.

` Scalability is still an important issue for database applications.

` Combininggclassification with database techniquesq should be a

promising research topic.

` Research Direction: Classification ofnon relational data e g

` Research Direction: Classification ofnon-relational data, e.g.,

text, spatial, multimedia, etc..

(11)

References

` C. Apte and S. Weiss. Data mining with decision trees and decision rules. Future Generation

C t S t 13 1997

Computer Systems, 13, 1997.

` L. Breiman, J. Friedman, R. Olshen, and C. Stone. Classification and Regression Trees. Wadsworth

International Group, 1984.p

` P. K. Chan and S. J. Stolfo. Learning arbiter and combiner trees from partitioned data for scaling

machine learning. In Proc. 1st Int. Conf. Knowledge Discovery and Data Mining (KDD'95), pages

39 44 M l C d A 1995

39-44, Montreal, Canada, August 1995.

` U. M. Fayyad. Branching on attribute values in decision tree generation. In Proc. 1994 AAAI Conf.,

pages 601-606, AAAI Press, 1994.

p g , ,

` J. Gehrke, R. Ramakrishnan, and V. Ganti. Rainforest: A framework for fast decision tree

construction of large datasets. In Proc. 1998 Int. Conf. Very Large Data Bases, pages 416-427, New York, NY, August 1998.

` M. Kamber, L. Winstone, W. Gong, S. Cheng, and J. Han. Generalization and decision tree induction:

Efficient classification in data mining. In Proc. 1997 Int. Workshop Research Issues on Data

Efficient classification in data mining. In Proc. 1997 Int. Workshop Research Issues on Data Engineering (RIDE'97), pages 111-120, Birmingham, England, April 1997.

41

References (2)

` J. Magidson. The chaid approach to segmentation modeling: Chi-squared automatic interaction detection. In R. P. Bagozzi, editor, Advanced Methods of Marketing Research, pages g , , g , p g 118-159. Blackwell Business, Cambridge Massechusetts, 1994.

` M. Mehta, R. Agrawal, and J. Rissanen. SLIQ : A fast scalable classifier for data mining. In Proc. 1996 Int. Conf. Extending Database Technology (EDBT'96), Avignon, France, March 1996.

` S. K. Murthy, Automatic Construction of Decision Trees from Data: A Multi-Diciplinary Survey, Data Mining and Knowledge Discovery 2(4): 345-389, 1998

` J. R. Quinlan. Bagging, boosting, and c4.5. In Proc. 13th Natl. Conf. on Artificial Intelligence J gg g g g (AAAI'96), 725-730, Portland, OR, Aug. 1996.

` R. Rastogi and K. Shim. Public: A decision tree classifer that integrates building and pruning. In Proc. 1998 Int. Conf. Very Large Data Bases, 404-415, New York, NY, August 1998.

` J. Shafer, R. Agrawal, and M. Mehta. SPRINT : A scalable parallel classifier for data mining. In Proc. 1996 Int. Conf. Very Large Data Bases, 544-555, Bombay, India, Sept. 1996.

` S. M. Weiss and C. A. Kulikowski. Computer Systems that Learn: Classification and Prediction Methods from Statistics, Neural Nets, Machine Learning, and Expert Systems. Morgan Kaufman, 1991.