2. - The Course
DS OLAP
DS DP DW DM
Association
DS Classification
DS = Data source Clustering
DW = Data warehouse
DM = Data Mining
DP = Staging Database
3. Chapter Objectives
Learn basic techniques for data classification
and prediction.
Realize the difference between the following
classifications of data:
– supervised classification
– prediction
– unsupervised classification
4. Chapter Outline
What is classification and prediction of data?
How do we classify data by decision tree induction?
What are neural networks and how can they classify?
What is Bayesian classification?
Are there other classification techniques?
How do we predict continuous values?
5. What is Classification?
The goal of data classification is to organize and
categorize data in distinct classes.
– A model is first created based on the data
distribution.
– The model is then used to classify new data.
– Given the model, a class can be predicted for new
data.
Classification = prediction for discrete and nominal
values
6. What is Prediction?
The goal of prediction is to forecast or deduce the value of an
attribute based on values of other attributes.
– A model is first created based on the data distribution.
– The model is then used to predict future or unknown values
In Data Mining
– If forecasting discrete value Classification
– If forecasting continuous value Prediction
7. Supervised and Unsupervised
Supervised Classification = Classification
– We know the class labels and the number of
classes
Unsupervised Classification = Clustering
– We do not know the class labels and may not
know the number of classes
8. Preparing Data Before
Classification
Data transformation:
– Discretization of continuous data
– Normalization to [-1..1] or [0..1]
Data Cleaning:
– Smoothing to reduce noise
Relevance Analysis:
– Feature selection to eliminate irrelevant attributes
9. Application
Credit approval
Target marketing
Medical diagnosis
Defective parts identification in manufacturing
Crime zoning
Treatment effectiveness analysis
Etc
10. Classification is a 3-step process
1. Model construction (Learning):
• Each tuple is assumed to belong to a predefined class, as
determined by one of the attributes, called the class label.
• The set of all tuples used for construction of the model is
called training set.
– The model is represented in the following forms:
• Classification rules, (IF-THEN statements),
• Decision tree
• Mathematical formulae
11. 1. Classification Process (Learning)
Name Income Age Credit
Samir Low <30
rating Classification Method
bad
Ahmed Medium [30...40
] good
Salah High <30 good
Ali Medium >40 good
Classification Model
Sami Low [30..40] good
Emad Medium <30 bad
IF Income = ‘High’
Training Data class OR Age > 30
THEN Class = ‘Good
OR
Decision Tree
OR
Mathematical For
12. Classification is a 3-step process
2. Model Evaluation (Accuracy):
– Estimate accuracy rate of the model based on a test set.
– 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 classified by the model.
– Test set is independent of training set otherwise over-fitting
will occur
13. 2. Classification Process (Accuracy
Evaluation)
Classification Model
Name Income Age Credit rating Model
Naser Low <30 Bad Bad
Accuracy
Lutfi Medium <30 Bad good 75%
Adel High >40 good good
Fahd Medium [30..40] good good
class
14. Classification is a three-step process
3. Model Use (Classification):
– The model is used to classify unseen objects.
• Give a class label to a new tuple
• Predict the value of an actual attribute
16. Classification Methods Classification Method
Decision Tree Induction
Neural Networks
Bayesian Classification
Association-Based Classification
K-Nearest Neighbour
Case-Based Reasoning
Genetic Algorithms
Rough Set Theory
Fuzzy Sets
Etc.
17. Evaluating Classification Methods
Predictive accuracy
– Ability of the model to correctly predict the class label
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
18. Chapter Outline
What is classification and prediction of data?
How do we classify data by decision tree induction ?
What are neural networks and how can they
classify?
What is Bayesian classification?
Are there other classification techniques?
How do we predict continuous values?
20. What is a Decision Tree?
A decision tree is a flow-chart-like tree structure.
– Internal node denotes 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
21. Sample Decision Tree
Excellent customers
Fair customers
80
Income
< 6K >= 6K
Age 50 No YES
20
2000 6000 10000
Income
22. Sample Decision Tree
80
Income
<6k >=6k
NO Age
Age 50 >=50
<50
NO Yes
20
2000 6000 10000
Income
23. Sample Decision Tree
Outlook Temp Humidity Windy Play?
sunny hot high FALSE No
sunny hot high TRUE No
overcast hot high FALSE Yes
rainy mild high FALSE Yes
rainy cool normal FALSE Yes
rainy cool Normal TRUE No
overcast cool Normal TRUE Yes
sunny mild High FALSE No
sunny cool Normal FALSE Yes
rainy mild Normal FALSE Yes
sunny mild normal TRUE Yes
overcast mild High TRUE Yes
overcast hot Normal FALSE Yes
rainy mild high TRUE No
http://www-lmmb.ncifcrf.gov/~toms/paper/primer/latex/index.html
http://directory.google.com/Top/Science/Math/Applications/Information_Theory/Papers/
24. Decision-Tree Classification Methods
The basic top-down decision tree generation
approach usually consists of two phases:
1. Tree construction
• At the start, all the training examples are at the root.
• Partition examples are recursively based on selected
attributes.
2. Tree pruning
• Aiming at removing tree branches that may reflect noise
in the training data and lead to errors when classifying
test data improve classification accuracy
25. How to Specify Test Condition?
Depends on attribute types
– Nominal
– Ordinal
– Continuous
Depends on number of ways to split
– 2-way split
– Multi-way split
26. Splitting Based on Nominal Attributes
Multi-way split: Use as many partitions as distinct
values.
CarType
Family Luxury
Sports
Binary split: Divides values into two subsets.
Need to find optimal partitioning.
CarType CarType
{Sports, OR {Family,
Luxury} {Family} Luxury} {Sports}
27. Splitting Based on Ordinal Attributes
Multi-way split: Use as many partitions as distinct
values.
Size
Small Large
Medium
Binary split: Divides values into two subsets.
Need to find optimal partitioning.
Size
Size {Medium,
{Small,
{Large}
OR Large} {Small}
Medium}
Size
{Small,
What about this split? Large} {Medium}
28. Splitting Based on Continuous Attributes
Different ways of handling
– Discretization to form an ordinal categorical
attribute
• Static – discretize once at the beginning
• Dynamic – ranges can be found by equal
interval bucketing, equal frequency bucketing
(percentiles), or clustering.
– Binary Decision: (A < v) or (A ≥ v)
• consider all possible splits and finds the best cut
• can be more compute intensive
30. Tree Induction
Greedy strategy.
– Split the records based on an attribute test that
optimizes certain criterion.
Issues
– Determine how to split the records
• How to specify the attribute test condition?
• How to determine the best split?
– Determine when to stop splitting
31. How to determine the Best Split
Good customers fair customers
Customers
Income Age
<10k >=10k young old
32. How to determine the Best Split
Greedy approach:
– Nodes with homogeneous class distribution are
preferred
Need a measure of node impurity:
High degree Low degree pure
of impurity of impurity
50% red 75% red 100% red
50% green 25% green 0% green
33. Measures of Node Impurity
Information gain
– Uses Entropy
Gain Ratio
– Uses Information
Gain and Splitinfo
Gini Index
– Used only for
binary splits
34. Algorithm for Decision Tree Induction
Basic algorithm (a greedy algorithm)
– Tree is constructed in a top-down recursive divide-and-conquer
manner
– At start, all the training examples are at the root
– Attributes are categorical (if continuous-valued, they are discretized
in advance)
– Examples are partitioned recursively based on selected attributes
– Test attributes are selected on the basis of a heuristic or statistical
measure (e.g., information gain)
Conditions for stopping partitioning
– All samples for a given node belong to the same class
– There are no remaining attributes for further partitioning – majority
voting is employed for classifying the leaf
– There are no samples left
36. Entropy: Used by ID3
Entropy(S) = - p log2 p - q log2 q
Entropy measures the impurity of S
S is a set of examples
p is the proportion of positive examples
q is the proportion of negative examples
37. ID3
outlook temperature humidity windy play play
sunny hot high FALSE no
sunny hot high TRUE no don’t play
overcast hot high FALSE yes
rainy mild high FALSE yes pno = 5/14
rainy cool normal FALSE yes
rainy cool normal TRUE no
overcast cool normal TRUE yes
sunny mild high FALSE no
sunny cool normal FALSE yes
rainy mild normal FALSE yes
sunny mild normal TRUE yes pyes = 9/14
overcast mild high TRUE yes
overcast hot normal FALSE yes
rainy mild high TRUE no
Impurity = - pyes log2 pyes - pno log2 pno
= - 9/14 log2 9/14 - 5/14 log2 5/14
= 0.94 bits
38. ID3 0.94 bits
play
don’t play
al play
xim tion 2
don't play play don't play
play don't play play don't play
ma ma
sunny 3
high 3 4
hot 2 2
FALSE 6 2
or
overcast 4 0 mild 4 2
infrainy ain 3
g 2
normal 6 1
cool 3 1
TRUE 3 3
outlook humidity temperature windy
sunny overcast rainy high normal hot mild cool false true
amount of information required to specify class of an example given that it reaches node
0.97 bits 0.0 bits 0.97 bits 0.98 bits 0.59 bits 1.0 bits 0.92 bits 0.81 bits 0.81 bits 1.0 bits
* 5/14 * 4/14 * 5/14 * 7/14 * 7/14 * 4/14 * 6/14 * 4/14 * 8/14 * 6/14
+ + + +
= 0.69 bits = 0.79 bits = 0.91 bits = 0.89 bits
gain: 0.25 bits gain: 0.15 bits gain: 0.03 bits gain: 0.05 bits
39. ID3 outlook play
don’t play
sunny overcast rainy
0.97 bits outlook
sunny
temperature
hot
humidity
high
windy
FALSE
play
no
sunny hot high TRUE no
sunny mild high FALSE no
sunny cool normal FALSE yes
al
xim tion sunny mild normal TRUE yes
ma ma
humidity
or
inf gain temperature windy
high normal hot mild cool false true
0.0 bits 0.0 bits 0.0 bits 1.0 bits 0.0 bits 0.92 bits 1.0 bits
* 3/5 * 2/5 * 2/5 * 2/5 * 1/5 * 3/5 * 2/5
+ + +
= 0.0 bits = 0.40 bits = 0.95 bits
gain: 0.97 bits gain: 0.57 bits gain: 0.02 bits
40. ID3 outlook
play
don’t play
outlook temperature humidity windy play
sunny overcast rainy rainy mild high FALSE yes
rainy cool normal FALSE yes
0.97 bits rainy
rainy
cool
mild
normal
normal
TRUE
FALSE
no
yes
rainy mild high TRUE no
humidity
humidity temperature windy
high normal
high normal hot mild cool false true
∅
1.0 bits 0.92 bits 0.92 bits 1.0 bits 0.0 bits 0.0 bits
*2/5 * 3/5 * 3/5 * 2/5 * 3/5 * 2/5
+ + +
= 0.95 bits = 0.95 bits = 0.0 bits
gain: 0.02 bits gain: 0.02 bits gain: 0.97 bits
41. ID3
outlook temperature humidity windy play
sunny hot high FALSE no
sunny hot high TRUE no
overcast hot high FALSE yes
rainy mild high FALSE yes
rainy cool normal FALSE yes
rainy cool normal TRUE no
overcast cool normal TRUE yes
sunny mild high FALSE no
sunny cool normal FALSE yes
rainy mild normal FALSE yes play
sunny mild normal TRUE yes
overcast
overcast
mild
hot
high
normal
TRUE
FALSE
yes
yes outlook don’t play
rainy mild high TRUE no
sunny overcast rainy
Yes
humidity windy
high
normal false true
No Yes Yes No
42. C4.5
Information gain measure is biased towards attributes with a large
number of values
C4.5 (a successor of ID3) uses gain ratio to overcome the problem
(normalization to information gain)
– GainRatio(A) = Gain(A)/SplitInfo(A)
v | Dj | | Dj |
SplitInfo A ( D ) = −∑ × log 2 ( )
j =1 |D| |D|
Ex.
5 5 4 4 5 5
SplitInfo A ( D) = − ×log 2 ( ) − ×log 2 ( ) − ×log 2 ( ) = 0.926
14 14 14 14 14 14
– gain_ratio(income) = 0.029/0.926 = 0.031
The attribute with the maximum gain ratio is selected as the
splitting attribute
43. CART
If a data set D contains examples from n classes, gini index,
gini(D) is defined as
n 2
gini( D) =1− ∑ p j
j =1
where pj is the relative frequency of class j in D
If a data set D is split on A into two subsets D1 and D2, the gini
index gini(D) is defined as
|D1| |D |
gini A ( D) = gini( D1) + 2 gini( D 2)
|D| |D|
Reduction in Impurity:
∆gini( A) = gini( D) − giniA ( D)
The attribute provides the smallest ginisplit(D) (or the largest
reduction in impurity) is chosen to split the node (need to
enumerate all the possible splitting points for each attribute)
44. CART
Ex. D has 9 tuples in buys_computer = “yes” and 5 in “no”
2 2
9 5
gini ( D) = 1 − − = 0.459
14 14
Suppose the attribute income partitions D into 10 in D1: {low,
medium} and 4 in D2
10 4
giniincome∈{low,medium} ( D ) = Gini ( D1 ) + Gini ( D1 )
14 14
but gini{medium,high} is 0.30 and thus the best since it is the lowest
All attributes are assumed continuous-valued
May need other tools, e.g., clustering, to get the possible split
values
Can be modified for categorical attributes
45. Comparing Attribute Selection Measures
The three measures, in general, return good results but
– Information gain:
• biased towards multivalued attributes
– Gain ratio:
• tends to prefer unbalanced splits in which one partition is
much smaller than the others
– Gini index:
• biased to multivalued attributes
• has difficulty when # of classes is large
• tends to favor tests that result in equal-sized partitions
and purity in both partitions
46. Other Attribute Selection Measures
CHAID: a popular decision tree algorithm, measure based on χ2 test for
independence
C-SEP: performs better than info. gain and gini index in certain cases
G-statistics: has a close approximation to χ2 distribution
MDL (Minimal Description Length) principle (i.e., the simplest solution
is preferred):
– The best tree as the one that requires the fewest # of bits to both
(1) encode the tree, and (2) encode the exceptions to the tree
Multivariate splits (partition based on multiple variable combinations)
– CART: finds multivariate splits based on a linear comb. of attrs.
Which attribute selection measure is the best?
– Most give good results, none is significantly superior than others
47. Underfitting and Overfitting
Overfitting
Underfitting: when model is too simple, both training and
test errors are large
49. Underfitting due to Insufficient Examples
Lack of data points in the lower half of the diagram makes it difficult
to predict correctly the class labels of that region
- Insufficient number of training records in the region causes the
decision tree to predict the test examples using other training
records that are irrelevant to the classification task
50. Two approaches to avoid Overfitting
Prepruning:
– Halt tree construction early—do not split a node if this would result
in the goodness measure falling below a threshold
– Difficult to choose an appropriate threshold
Postpruning:
– Remove branches from a “fully grown” tree—get a sequence of
progressively pruned trees
– Use a set of data different from the training data to decide
which is the “best pruned tree”
51. Scalable Decision Tree Induction Methods
ID3, C4.5, and CART are not efficient when the training set
doesn’t fit the available memory. Instead the following algorithms
are used
– SLIQ
• Builds an index for each attribute and only class list and
the current attribute list reside in memory
– SPRINT
• Constructs an attribute list data structure
– RainForest
• Builds an AVC-list (attribute, value, class label)
– BOAT
• Uses bootstrapping to create several small samples
52. BOAT
BOAT (Bootstrapped Optimistic Algorithm for Tree
Construction)
– Use a statistical technique called bootstrapping to create several
smaller samples (subsets), each fits in memory
– Each subset is used to create a tree, resulting in several trees
– These trees are examined and used to construct a new tree T’
• It turns out that T’ is very close to the tree that would be
generated using the whole data set together
– Adv: requires only two scans of DB, an incremental alg.
53. Why decision tree induction in data mining?
Relatively faster learning speed (than other
classification methods)
Convertible to simple and easy to understand
classification rules
Comparable classification accuracy with other
methods
54. Converting Tree to Rules
Outlook
Sunny Overcast Rain
Humidity Yes Wind
High Normal Strong Weak
No Yes No Yes
R1: IF (Outlook=Sunny) AND (Humidity=High) THEN Play=No
R2: IF (Outlook=Sunny) AND (Humidity=Normal) THEN Play=Yes
R3: IF (Outlook=Overcast) THEN Play=Yes
R4: IF (Outlook=Rain) AND (Wind=Strong) THEN Play=No
R5: IF (Outlook=Rain) AND (Wind=Weak) THEN Play=Yes
57. Basic Statistics
Assume
• D = All students
• X = ICS students
• C = SWE students 74 D
X 6 4 C
16
|X| = 10 P(X) = 10/100 P(X|C) = P(X,C)/P(C) = 4/20
|C| = 20 P(C) = 20/100 P(C|X) = P(X,C)/P(X) = 4/10
|D| = 100 P(X,C) = 4/100
P(X,C) = P(C|X)*P(X) = P(X|C)*P(C)
58. Bayesian Classifier – Basic Equation
P(X,C) = P(C|X)*P(X) = P(X|C)*P(C)
Class Prior Probability Descriptor Posterior Probability
P( C ) P( X | C )
P( C | X ) =
P( X )
Class Posterior Probability
Descriptor Prior Probability
59. Naive Bayesian Classifier
P ( C | X ) = P( C ) P( X | C )
P( X )
P (C1 )
P( C1 | X ) = P( x1 | C1 ) P( x2 | C1 ) P( x3 | C1 ) .... P( xn | C1 )
P(X)
P(C2 )
P( C2 | X ) = P( x1 | C2 ) P( x2 | C2 ) P( x3 | C2 ) .... P( xn | C2 )
P( X)
P(Cm )
P( Cm | X ) = P( x1 | Cm ) P( x2 | Cm ) P( x3 | Cm ) .... P( xn | Cm )
P( X)
Independence assumption about descriptors
60. Training Data
Outlook Temp Humidity Windy Play?
sunny hot high FALSE No
sunny hot high TRUE No
overcast hot high FALSE Yes
rainy mild high FALSE Yes
rainy cool normal FALSE Yes
rainy cool Normal TRUE No
overcast cool Normal TRUE Yes
sunny mild High FALSE No
sunny cool Normal FALSE Yes
rainy mild Normal FALSE Yes
sunny mild normal TRUE Yes
overcast mild High TRUE Yes
overcast hot Normal FALSE Yes
rainy mild high TRUE No
P(yes) = 9/14
P(no) = 5/14
61. Bayesian Classifier – Probabilities for the weather data
Frequency Tables
Outlook | No Yes Temp. | No Yes Humidity | No Yes Windy | No Yes
---------------------------------- ---------------------------------- ---------------------------------- ----------------------------------
Sunny | 3 2 Hot | 2 2 High | 4 3 False | 2 6
---------------------------------- ---------------------------------- ---------------------------------- ----------------------------------
Overcast | 0 4 Mild | 2 4 Normal | 1 6 True | 3 3
---------------------------------- ----------------------------------
Rainy | 2 3 Cool | 1 3
Outlook | No Yes Temp. | No Yes Humidity | No Yes Windy | No Yes
---------------------------------- ---------------------------------- ---------------------------------- ----------------------------------
Sunny | 3/5 2/9 Hot | 2/5 2/9 High | 4/5 3/9 False | 2/5 6/9
---------------------------------- ---------------------------------- ---------------------------------- ----------------------------------
Overcast | 0/5 4/9 Mild | 2/5 4/9 Normal | 1/5 6/9 True | 3/5 3/9
---------------------------------- ----------------------------------
Rainy | 2/5 3/9 Cool | 1/5 3/9
Likelihood Tables
62. Bayesian Classifier – Predicting a new day
Outlook Temp. Humidity Windy Play
X sunny cool high true ? Class?
P(yes|X) = p(sunny|yes) x p(cool|yes) x p(high|yes) x p(true|yes) x p(yes)
= 2/9 x 3/9 x 3/9 x 3/9 x 9/14 = 0.0053 => 0.0053/(0.0053+0.0206) = 0.205
P(no|X) = p(sunny|no) x p(cool|no) x p(high|no) x p(true|no) x p(no)
= 3/5 x 1/5 x 4/5 x 3/5 x 5/14 = 0.0206=0.0206/(0.0053+0.0206) = 0.795
63. Bayesian Classifier – zero frequency problem
What if a descriptor value doesn’t occur with every class value
P(outlook=overcast|No)=0
Remedy: add 1 to the count for every descriptor-class combination
(Laplace Estimator)
Outlook | No Yes Temp. | No Yes Humidity | No Yes Windy | No Yes
---------------------------------- ---------------------------------- ---------------------------------- ----------------------------------
Sunny | 3+1 2+1 Hot | 2+1 2+1 High | 4+1 3+1 False | 2+1 6+1
---------------------------------- ---------------------------------- ---------------------------------- ----------------------------------
Overcast | 0+1 4+1 Mild | 2+1 4+1 Normal | 1+1 6+1 True | 3+1 3+1
---------------------------------- ----------------------------------
Rainy | 2+1 3+1 Cool | 1+1 3+1
64. Bayesian Classifier – General Equation
P ( X | Ck ) P( Ck )
P ( Ck | X ) =
P( X )
Likelihood: P ( X | Ck )
1 ( x − µ )2
Continues variable: P ( x | C ) = exp−
(2πσ )
2 1/ 2
2σ 2
67. Naïve Bayesian Classifier: Comments
Advantages
– Easy to implement
– Good results obtained in most of the cases
Disadvantages
– Assumption: class conditional independence, therefore loss of
accuracy
– Practically, dependencies exist among variables
• E.g., hospitals: patients: Profile: age, family history, etc.
Symptoms: fever, cough etc., Disease: lung cancer,
diabetes, etc.
• Dependencies among these cannot be modeled by Naïve
Bayesian Classifier
How to deal with these dependencies?
– Bayesian Belief Networks
68. Bayesian Belief Networks
Bayesian belief network allows a subset of the variables
conditionally independent
A graphical model of causal relationships
– Represents dependency among the variables
– Gives a specification of joint probability distribution
Nodes: random variables
Links: dependency
X Y X and Y are the parents of Z, and Y is
the parent of P
Z No dependency between Z and P
P
Has no loops or cycles
69. Bayesian Belief Network: An Example
The conditional probability table
Family (CPT) for variable LungCancer:
Smoker
History
(FH, S) (FH, ~S) (~FH, S) (~FH, ~S)
LC 0.8 0.5 0.7 0.1
~LC 0.2 0.5 0.3 0.9
LungCancer Emphysema
CPT shows the conditional probability for
each possible combination of its parents
PositiveXRay Dyspnea Derivation of the probability of a
particular combination of values of X,
from CPT:
n
Bayesian Belief Networks P ( x1 ,..., xn ) = ∏ P ( x i | Parents (Y i ))
i =1
70. Training Bayesian Networks
Several scenarios:
– Given both the network structure and all variables
observable: learn only the CPTs
– Network structure known, some hidden variables: gradient
descent (greedy hill-climbing) method, analogous to neural
network learning
– Network structure unknown, all variables observable:
search through the model space to reconstruct network
topology
– Unknown structure, all hidden variables: No good
algorithms known for this purpose.
81. Support Vector Machines
w• x + b = 0
w • x + b = +1
w • x + b = −1
1 if w • x + b ≥ 1 2
f ( x) = Margin = 2
−1 if w • x + b ≤ −1 || w ||
82. Finding the Decision Boundary
Let {x1, ..., xn} be our data set and let yi ∈ {1,-1} be the class
label of xi
The decision boundary should classify all points correctly ⇒
The decision boundary can be found by solving the following
constrained optimization problem
This is a constrained optimization problem. Solving it is beyond
our course
83. Support Vector Machines
2
We want to maximize: Margin = 2
|| w ||
2
|| w ||
– Which is equivalent to minimizing: L( w) =
2
– But subjected to the following constraints:
1 if w • x i + b ≥ 1
f ( xi ) =
−1 if w • x i + b ≤ −1
• This is a constrained optimization problem
– Numerical approaches to solve it (e.g., quadratic
programming)
84. Classifying new Tuples
The decision boundary is determined only by the support vectors
Let tj (j=1, ..., s) be the indices of the s support vectors.
For testing with a new data z
– Compute and
classify z as class 1 if the sum is positive, and class 2
otherwise
86. Support Vector Machines
What if the training set is not linearly separable?
Slack variables ξi can be added to allow misclassification of
difficult or noisy examples, resulting margin called soft.
ξi
ξi
87. Support Vector Machines
What if the problem is not linearly separable?
– Introduce slack variables
• Need to minimize:
2
|| w || N
k
L( w) = + C ∑ ξi
2 i =1
• Subject to:
1 if w • x i + b ≥ 1 - ξi
f ( xi ) =
−1 if w • x i + b ≤ −1 + ξi
89. Non-linear SVMs
Datasets that are linearly separable with some noise work out
great:
0 x
But what are we going to do if the dataset is just too hard?
0 x
How about… mapping data to a higher-dimensional space:
x2
0 x
90. Non-linear SVMs: Feature spaces
General idea: the original feature space can always be mapped to
some higher-dimensional feature space where the training set is
separable:
Φ: x → φ(x)
92. What Is Prediction?
(Numerical) prediction is similar to classification
– construct a model
– use model to predict continuous or ordered value for a given
input
Prediction is different from classification
– Classification refers to predict categorical class label
– Prediction models continuous-valued functions
Major method for prediction: regression
– model the relationship between one or more predictor
variables and a response variable
95. Regression Analysis
Simple Linear regression
multiple regression
Non-linear regression
Other regression methods:
– generalized linear model,
– Poisson regression,
– log-linear models,
– regression trees
96. Simple Linear Regression
describes the linear relationship between a predictor variable,
plotted on the x-axis, and a response variable, plotted on the
y-axis
Y
X
102. Least Squares Regression
ˆ
Model line: Y = β 0 + β1 X
Residual (ε) = Y − Yˆ
Sum of squares of residuals = ∑ ˆ
(Y − Y ) 2
we must find values of β o and β1 that minimise
∑ ˆ
(Y − Y ) 2
103. Linear Regression
A model line: y = w0 + w1 x acquired by using Method
of least squares to estimates the best-fitting straight
line has:
w = y−w x
0 1
| D|
∑( x − x )( yi − y )
w =
i
i=1
1 ∑( x
| D|
i − x )2
i=1
104. Multiple Linear Regression
Multiple linear regression: involves more than one predictor
variable
The linear model with a single predictor variable X can easily
be extended to two or more predictor variables
Y = β o + β1 X 1 + β 2 X 2 + ... + β p X p + ε
– Solvable by extension of least square method or using SAS,
S-Plus
105. Nonlinear Regression
Some nonlinear models can be modeled by a polynomial
function
A polynomial regression model can be transformed into linear
regression model. For example,
y = w0 + w1 x + w2 x2 + w3 x3
convertible to linear with new variables: x2 = x2, x3= x3
y = w0 + w1 x + w2 x2 + w3 x3
Other functions, such as power function, can also be
transformed to linear model
Some models are intractable nonlinear
– possible to obtain least square estimates through extensive
calculation on more complex formulae
107. What is a ANN?
ANN is a data structure that supposedly simulates
the behavior of neurons in a biological brain.
ANN is composed of layers of units interconnected.
Messages are passed along the connections from
one unit to the other.
Messages can change based on the weight of the
connection and the value in the node
109. ANN
Output Y is 1 if at least two of the three inputs are equal to 1.
110. ANN
Y = I (0.3 X 1 + 0.3 X 2 + 0.3 X 3 − 0.4 > 0)
1 if z is true
where I ( z ) =
0 otherwise
111. Artificial Neural Networks
Model is an assembly of
inter-connected nodes and
weighted links
Output node sums up each
of its input value according
to the weights of its links
Perceptron Model
Compare output node Y = I ( ∑wi X i − t ) or
against some threshold t i
Y = sign( ∑ wi X i − t )
i
112. Neural Networks
Advantages
– prediction accuracy is generally high.
– robust, works when training examples contain errors.
– output may be discrete, real-valued, or a vector of several
discrete or real-valued attributes.
– fast evaluation of the learned target function.
Criticism
– long training time.
– difficult to understand the learned function (weights).
– not easy to incorporate domain knowledge.
113. Learning Algorithms
Back propagation for classification
Kohonen feature maps for clustering
Recurrent back propagation for classification
Radial basis function for classification
Adaptive resonance theory
Probabilistic neural networks
114. Major Steps for Back Propagation
Network
Constructing a network
– input data representation
– selection of number of layers, number of nodes in
each layer.
Training the network using training data
Pruning the network
Interpret the results
116. How A Multi-Layer Neural Network Works?
The inputs to the network correspond to the attributes measured for
each training tuple
Inputs are fed simultaneously into the units making up the input layer
They are then weighted and fed simultaneously to a hidden layer
The number of hidden layers is arbitrary, although usually only one
The weighted outputs of the last hidden layer are input to units making
up the output layer, which emits the network's prediction
The network is feed-forward in that none of the weights cycles back to
an input unit or to an output unit of a previous layer
From a statistical point of view, networks perform nonlinear
regression: Given enough hidden units and enough training samples,
they can closely approximate any function
117. Defining a Network Topology
First decide the network topology: # of units in the input layer,
# of hidden layers (if > 1), # of units in each hidden layer, and #
of units in the output layer
Normalizing the input values for each attribute measured in the
training tuples to [0.0—1.0]
One input unit per domain value
Output, if for classification and more than two classes, one
output unit per class is used
Once a network has been trained and its accuracy is
unacceptable, repeat the training process with a different
network topology or a different set of initial weights
118. Backpropagation
Iteratively process a set of training tuples & compare the network's
prediction with the actual known target value
For each training tuple, the weights are modified to minimize the
mean squared error between the network's prediction and the
actual target value
Modifications are made in the “backwards” direction: from the
output layer, through each hidden layer down to the first hidden
layer, hence “backpropagation”
Steps
– Initialize weights (to small random #s) and biases in the network
– Propagate the inputs forward (by applying activation function)
– Backpropagate the error (by updating weights and biases)
– Terminating condition (when error is very small, etc.)
119. Backpropagation
Err j = O j (1 − O j )∑ Errk w jk
k
wij = wij + (l ) Err j Oi
θ j = θ j + (l) Err j
Err j = O j (1 − O j )(T j − O j )
Generated value Correct value
120. Network Pruning
Fully connected network will be hard to articulate
n input nodes, h hidden nodes and m output nodes
lead to h(m+n) links (weights)
Pruning: Remove some of the links without affecting
classification accuracy of the network.
121. Other Classification Methods
Associative classification : Association rule based condSet
class
Genetic algorithm : Initial population of encoded rules are
changed by mutation and cross-over based on survival of
accurate once (survival).
K-nearest neighbor classifier : Learning by analogy.
Case-based reasoning : Similarity with other cases.
Rough set theory : Approximation to equivalence classes.
Fuzzy sets: Based on fuzzy logic (truth values between 0..1).
123. Lazy vs. Eager Learning
Lazy vs. eager learning
– Lazy learning (e.g., instance-based learning): Simply
stores training data (or only minor processing) and waits
until it is given a test tuple
– Eager learning (the above discussed methods): Given a
set of training set, constructs a classification model
before receiving new (e.g., test) data to classify
Lazy: less time in training but more time in predicting
124. Lazy Learner: Instance-Based Methods
Instance-based learning:
– Store training examples and delay the processing (“lazy
evaluation”) until a new instance must be classified
Typical approaches
– k-nearest neighbor approach
• Instances represented as points in a Euclidean
space.
– Case-based reasoning
• Uses symbolic representations and knowledge-
based inference
125. Nearest Neighbor Classifiers
Basic idea:
– If it walks like a duck, quacks like a duck, then it’s
probably a duck
Compute
Distance Test
Record
Choose k of the
“nearest” records
Training
records
126. Instance-Based Classifiers
• Store the training records
• Use training records to
predict the class label of
unseen cases
127. Definition of Nearest Neighbor
X X X
(a) 1-nearest neighbor (b) 2-nearest neighbor (c) 3-nearest neighbor
K-nearest neighbors of a record x are data points
that have the k smallest distance to x
128. The k-Nearest Neighbor Algorithm
All instances correspond to points in the n-D space
The nearest neighbor are defined in terms of Euclidean
distance, dist(X1, X2)
Target function could be discrete- or real- valued
For discrete-valued, k-NN returns the most common value
among the k training examples nearest to xq
Vonoroi diagram: the decision surface induced by 1-NN for a
typical set of training examples
_
_ _
.
_
+
_ .
+
+
. . .
xq
_ + .
129. Nearest-Neighbor Classifiers
Requires three things
– The set of stored records
– Distance Metric to compute
distance between records
– The value of k, the number of
nearest neighbors to retrieve
To classify an unknown record:
– Compute distance to other training
records
– Identify k nearest neighbors
– Use class labels of nearest
neighbors to determine the class
label of unknown record (e.g., by
taking majority vote)
130. Nearest Neighbor Classification
Compute distance between two points:
– Euclidean distance
d ( p, q ) = ∑( p i
i
−q )
i
2
Determine the class from nearest neighbor list
– take the majority vote of class labels among the k-
nearest neighbors
– Weigh the vote according to distance
• weight factor, w = 1/d2
131. Nearest Neighbor Classification…
Scaling issues
– Attributes may have to be scaled to prevent
distance measures from being dominated by one of
the attributes
– Example:
• height of a person may vary from 1.5m to 1.8m
• weight of a person may vary from 90lb to 300lb
• income of a person may vary from $10K to $1M
132. Nearest Neighbor Classification…
Choosing the value of k:
– If k is too small, sensitive to noise points
– If k is too large, neighborhood may include points from other
classes
133. Metrics for Performance Evaluation
Focus on the predictive capability of a model
– Rather than how fast it takes to classify or build models,
scalability, etc.
Confusion Matrix:
PREDICTED CLASS a: TP (true positive)
Class=Yes Class=No b: FN (false negative)
c: FP (false positive)
Class=Yes a b
ACTUAL d: TN (true negative)
CLASS Class=No c d
134. Metrics for Performance Evaluation…
PREDICTED CLASS
Class=Yes Class=No
ACTUAL Class=Yes a b
CLASS (TP) (FN)
Class=No c d
(FP) (TN)
Most widely-used metric:
a+d TP + TN
Accuracy = =
a + b + c + d TP + TN + FP + FN
Error Rate = 1 - Accuracy
135. Limitation of Accuracy
Consider a 2-class problem
– Number of Class 0 examples = 9990
– Number of Class 1 examples = 10
If model predicts everything to be class 0, accuracy is
9990/10000 = 99.9 %
– Accuracy is misleading because model does not
detect any class 1 example
137. Predictor Error Measures
Test error (generalization error): the average loss over the test set
d
– Mean absolute error: ∑| yi − yi ' |
i =1
d
d
– Mean squared error: ∑(y
i =1
i − yi ' ) 2
d
d
∑y
| i −yi ' |
– Relative absolute error: i=
d
1
∑y
|
i=1
i −y |
d
∑(y
i =1
i − yi ' ) 2
– Relative squared error: d
∑(y
i =1
i − y)2
– The mean squared-error exaggerates the presence of outliers
Popularly use (square) root mean-square error, similarly, root
relative squared error
138. Evaluating Accuracy
Holdout method
– Given data is randomly partitioned into two independent sets
• Training set (e.g., 2/3) for model construction
• Test set (e.g., 1/3) for accuracy estimation
– Random sampling: a variation of holdout
• Repeat holdout k times, accuracy = avg. of the
accuracies obtained
Cross-validation (k-fold, where k = 10 is most popular)
– Randomly partition the data into k mutually exclusive
subsets, each approximately equal size
– At i-th iteration, use Di as test set and others as training set
139. Evaluating Accuracy
Bootstrap
– Works well with small data sets
– Samples the given training tuples uniformly with replacement
Several boostrap methods, and a common one is .632 boostrap
– Suppose we are given a data set of d tuples. The data set is sampled
d times, with replacement, resulting in a training set of d samples. The
data tuples that did not make it into the training set end up forming the
test set. About 63.2% of the original data will end up in the bootstrap,
and the remaining 36.8% will form the test set (since (1 – 1/d)d ≈ e-1 =
0.368)
– Repeat the sampling procedure k times, overall accuracy of the model:
k
acc( M ) = ∑ (0.632 × acc( M i ) test _ set +0.368 × acc( M i ) train _ set )
i =1
140. Ensemble Methods
Construct a set of classifiers from the training data
Predict class label of previously unseen records by
aggregating predictions made by multiple classifiers
– Use a combination of models to increase accuracy
– Combine a series of k learned models, M1, M2, …, Mk, with the aim
of creating an improved model M*
Popular ensemble methods
– Bagging
• averaging the prediction over a collection of classifiers
– Boosting
• weighted vote with a collection of classifiers
142. Bagging: Boostrap Aggregation
Analogy: Diagnosis based on multiple doctors’ majority vote
Training
– Given a set D of d tuples, at each iteration i, a training set Di of d
tuples is sampled with replacement from D (i.e., boostrap)
– A classifier model Mi is learned for each training set Di
Classification: classify an unknown sample X
– Each classifier Mi returns its class prediction
– The bagged classifier M* counts the votes and assigns the class
with the most votes to X
Prediction: can be applied to the prediction of continuous values
by taking the average value of each prediction for a given test
tuple
143. Bagging: Boostrap Aggregation
Accuracy
– Often significant better than a single classifier derived
from D
– For noise data: not considerably worse, more robust
– Proved improved accuracy in prediction
144. Boosting
Analogy: Consult several doctors, based on a combination of
weighted diagnoses—weight assigned based on the previous
diagnosis accuracy
How boosting works?
– Weights are assigned to each training tuple
– A series of k classifiers is iteratively learned
– After a classifier Mi is learned, the weights are updated to
allow the subsequent classifier, Mi+1, to pay more attention to
the training tuples that were misclassified by Mi
– The final M* combines the votes of each individual classifier,
where the weight of each classifier's vote is a function of its
accuracy
145. Boosting
The boosting algorithm can be extended for the
prediction of continuous values
Comparing with bagging: boosting tends to achieve
greater accuracy, but it also risks overfitting the
model to misclassified data
146. Boosting: Adaboost
Given a set of d class-labeled tuples, (X1, y1), …, (Xd, yd)
Initially, all the weights of tuples are set the same (1/d)
Generate k classifiers in k rounds. At round i,
– Tuples from D are sampled (with replacement) to form a training set
Di of the same size
– Each tuple’s chance of being selected is based on its weight
– A classification model Mi is derived from Di
– Its error rate is calculated using Di as a test set
– If a tuple is misclassified, its weight is increased, otherwise it is
decreased
Error rate: err(Xj) is the misclassification error of tuple Xj. Classifier
Mi error rate is the sum of the weights of the misclassified tuples:
d
error ( M i ) = ∑ j ×err ( X j )
w
j
1 − error ( M i )
log
error ( M i )
The weight of classifier Mi’s vote is