Classification and

Prediction
 Classification and Prediction
 What is classification? What is
regression?
 Issues regarding classification and
prediction
 Classification by decision tree induction
 Scalable decision tree induction
 Classification vs. Prediction
 Classification:
◼ predicts categorical class labels
◼ classifies data (constructs a model) based on the
training set and the values (class labels) in a classifying
attribute and uses it in classifying new data
 Regression:
◼ models continuous-valued functions, i.e., predicts
unknown or missing values
 Typical Applications
◼ credit approval
◼ target marketing
◼ medical diagnosis
◼ treatment effectiveness analysis
 Why Classification? A motivating
application
 Credit approval
◼ A bank wants to classify its customers based on whether
they are expected to pay back their approved loans
◼ The history of past customers is used to train the
classifier
◼ The classifier provides rules, which identify potentially
reliable future customers
◼ Classification rule:
 If age = “31...40” and income = high then credit_rating =
excellent
◼ Future customers
 Paul: age = 35, income = high  excellent credit rating
 John: age = 20, income = medium  fair credit rating
Classification—A Two-Step Process
 Model construction: describing a set of predetermined
classes
◼ Each tuple/sample is assumed to belong to a predefined class,
as determined by the class label attribute
◼ The set of tuples used for model construction: training set
◼ The model is represented as classification rules, decision
trees, or mathematical formulae
 Model usage: for classifying future or unknown objects
◼ Estimate accuracy of the model
 The known label of test samples 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
 Classification Process (1):
Model Construction
Classification
Algorithms
Training
Data

NAME RANK YEARS TENURED Classifier

M ike A ssistant P rof 3 no (Model)
M ary A ssistant P rof 7 yes
B ill P rofessor 2 yes
Jim A ssociate P rof 7 yes
IF rank = ‘professor’
D ave A ssistant P rof 6 no
OR years > 6
A nne A ssociate P rof 3 no
THEN tenured = ‘yes’
 Classification Process (2): Use
the Model in Prediction
Accuracy=?
Classifier

Testing
Data Unseen Data

(Jeff, Professor, 4)
NAME RANK YEARS TENURED
Tom Assistant Prof 2 no Tenured?
Mellisa Associate Prof 7 no
George Professor 5 yes
Joseph Assistant Prof 7 yes
Supervised vs. Unsupervised
Learning
 Supervised learning (classification)
◼ Supervision: The training data (observations,
measurements, etc.) are accompanied by labels
indicating the class of the observations
◼ New data is classified based on the training set
 Unsupervised learning (clustering)
◼ The class labels of training data is unknown
◼ Given a set of measurements, observations, etc. with
the aim of establishing the existence of classes or
clusters in the data
Issues regarding classification and
prediction (1): Data Preparation
 Data cleaning
◼ Preprocess data in order to reduce noise and handle
missing values
 Relevance analysis (feature selection)
◼ Remove the irrelevant or redundant attributes
 Data transformation
◼ Generalize and/or normalize data
 numerical attribute income  categorical
{low,medium,high}
 normalize all numerical attributes to [0,1)
Issues regarding classification and prediction
(2): Evaluating Classification Methods
 Predictive accuracy
 Speed
◼ time to construct the model
◼ time to use the model
 Robustness
◼ handling noise and missing values
 Scalability
◼ efficiency in disk-resident databases
 Interpretability:
◼ understanding and insight provided by the model
 Goodness of rules (quality)
◼ decision tree size
◼ compactness of classification rules
 Classification by Decision Tree
Induction
 Decision tree
◼ A flow-chart-like tree structure
◼ Internal node denotes a test on an attribute
◼ Branch represents an outcome of the test
◼ Leaf nodes represent class labels or class distribution
 Decision tree generation consists of two phases
◼ Tree construction
 At start, all the training examples are at the root

 Partition examples recursively based on selected attributes

◼ Tree pruning
 Identify and remove branches that reflect noise or outliers

 Use of decision tree: Classifying an unknown sample

◼ Test the attribute values of the sample against the decision tree
 Training Dataset
age income student credit_rating buys_computer
This <=30 high no fair no

follows
<=30 high no excellent no
31…40 high no fair yes
an >40 medium no fair yes
example >40 low yes fair yes
>40 low yes excellent no
from 31…40 low yes excellent yes
Quinlan’s <=30 medium no fair no
ID3 <=30
>40
low
medium
yes fair
yes fair
yes
yes
<=30 medium yes excellent yes
31…40 medium no excellent yes
31…40 high yes fair yes
>40 medium no excellent no
Output: A Decision Tree for
“buys_computer”

age?

<=30 overcast
30..40 >40

student? yes credit rating?

no yes excellent fair

no yes no yes
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)
◼ Samples 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
Algorithm for Decision Tree
Induction (pseudocode)
Algorithm GenDecTree(Sample S, Attlist A)
1. create a node N
2. If all samples are of the same class C then label N with C;
terminate;
3. If A is empty then label N with the most common class C in
S (majority voting); terminate;
4. Select aA, with the highest information gain; Label N with
a;
5. For each value v of a:
a. Grow a branch from N with condition a=v;
b. Let Sv be the subset of samples in S with a=v;
c. If Sv is empty then attach a leaf labeled with the most
common class in S;
d. Else attach the node generated by GenDecTree(Sv, A-a)
 Attribute Selection Measure:
Information Gain (ID3/C4.5)
◼ Select the attribute with the highest information gain
◼ Let pi be the probability that an arbitrary tuple in D
belongs to class Ci, estimated by |Ci, D|/|D|
◼ Expected information (entropy) needed to classify a tuple
in D: m
Info( D ) = − pi log 2 ( pi )
i =1

◼ Information needed (after using A to split D into v

partitions) to classify D: v |D |
InfoA ( D) =   I (D j )
j

j =1 | D |

◼ Information gained by branching on attribute A

Gain(A) = Info(D) − InfoA(D)
 Attribute Selection: Information Gain
 Class P: buys_computer = “yes”
 Class N: buys_computer = “no”
9 9 5 5
Info( D) = I (9,5) = − log 2 ( ) − log 2 ( ) =0.940
14 14 14 14
age pi ni I(pi, ni) 5 4
Infoage ( D) = I (2,3) + I (4,0)
<=30 2 3 0.971 14 14
31…40 4 0 0 5
+ I (3,2) = 0.694
>40 3 2 0.971 14

age income student credit_rating buys_computer Gain(age) = Info( D) − Infoage ( D) = 0.246

<=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
31…40
low
low
yes excellent
yes excellent
no
yes Gain(income) = 0.029
Gain( student ) = 0.151
<=30 medium no fair no
<=30 low yes fair yes
>40 medium yes fair yes
<=30
31…40
medium
medium
yes excellent
no excellent
yes
yes
Gain(credit _ rating ) = 0.048
31…40 high yes fair yes
>40 medium no excellent no
Splitting the samples using age
age?
<=30 >40
30...40
income student credit_rating buys_computer income student credit_rating buys_computer
high no fair no medium no fair yes
high no excellent no low yes fair yes
medium no fair no low yes excellent no
low yes fair yes medium yes fair yes
medium yes excellent yes medium no excellent no

income student credit_rating buys_computer

high no fair yes
low yes excellent yes labeled yes
medium no excellent yes
high yes fair yes
 Computing Information-Gain for
Continuous-Value Attributes
 Let attribute A be a continuous-valued attribute
 Must determine the best split point for A
◼ Sort the value A in increasing order
◼ Typically, the midpoint between each pair of adjacent
values is considered as a possible split point
 (ai+ai+1)/2 is the midpoint between the values of a i and ai+1
◼ The point with the minimum expected information
requirement for A is selected as the split-point for A
 Split:
◼ D1 is the set of tuples in D satisfying A ≤ split-point, and
D2 is the set of tuples in D satisfying A > split-point
Gain Ratio for Attribute Selection (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) SplitInfoA ( D) = − | D j |  log 2 ( | D j | )
v

j =1 |D| |D|

4 4 6 6 4 4
SplitInfoA ( D) = −  log 2 ( ) −  log 2 ( ) −  log 2 ( ) = 0.926
14 14 14 14 14 14
◼ GainRatio(A) = Gain(A)/SplitInfo(A)
 Ex. gain_ratio(income) = 0.029/0.926 = 0.031
 The attribute with the maximum gain ratio is
selected as the splitting attribute
Gini index (CART, IBM
IntelligentMiner)
 If a data set D contains examples from n classes, gini index,
gini(D) is defined as n
gini( D) = 1−  p 2j
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(D2)
|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)
Gini index (CART, IBM
IntelligentMiner)
 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 gini  10  4
( D) =  Gini( D ) +  Gini( D )
income{low , medium} 1 1
 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
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
 Comparison among Splitting Criteria
For a 2-class problem:
Overfitting and Tree Pruning
 Overfitting: An induced tree may overfit the
training data
◼ Too many branches, some may reflect anomalies due to noise or
outliers
◼ Poor accuracy for unseen samples

 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”
Classification in Large Databases
 Classification—a classical problem extensively
studied by statisticians and machine learning
researchers
 Scalability: Classifying data sets with millions of
examples and hundreds of attributes with
reasonable speed
 Why decision tree induction in data mining?
◼ relatively faster learning speed (than other classification
methods)
◼ convertible to simple and easy to understand classification
rules
◼ can use SQL queries for accessing databases
◼ comparable classification accuracy with other methods
Scalable Decision Tree Induction Methods

 SLIQ (EDBT’96 — Mehta et al.)

◼ Builds an index for each attribute and only class list and
the current attribute list reside in memory
 SPRINT (VLDB’96 — J. Shafer et al.)
◼ Constructs an attribute list data structure
 PUBLIC (VLDB’98 — Rastogi & Shim)
◼ Integrates tree splitting and tree pruning: stop growing the
tree earlier
 RainForest (VLDB’98 — Gehrke, Ramakrishnan &
Ganti)
◼ Builds an AVC-list (attribute, value, class label)
 BOAT (PODS’99 — Gehrke, Ganti, Ramakrishnan &
Loh)
◼ Uses bootstrapping to create several small samples
 SLIQ (Supervised Learning In Quest)

 Decision-tree classifier for data mining

 Design goals:
◼ Able to handle large disk-resident training
sets
◼ No restrictions on training-set size
Building tree
GrowTree(TrainingData D)
Partition(D);

Partition(Data D)
if (all points in D belong to the same class) then
return;
for each attribute A do
evaluate splits on attribute A;
use best split found to partition D into D1 and D2;
Partition(D1);
Partition(D2);
 Data Setup

 One list for each attribute

 Entries in an Attribute List consist of:
◼ attribute value
◼ class list index
 A list for the classes with pointers to the tree nodes

 Lists for continuous attributes are in sorted order

 Attribute lists may be disk-resident
 Class List must be in main memory
 Data Setup
Attribute lists Class list
Age CLI Car Type CLI Risk Leaf
23 0 family 0 High N1
Age Car Type Risk
17 1 sports 1 High N1
23 family High 43 2 sports 2
17 sports High High N1
68 3 family 3
43 sports High 32 4 truck 4
Low N1
68 family Low 20 5 family 5 Low N1
32 truck Low High N1
20 family High

Age CLI Car Type CLI Risk Leaf

17 1 family 0 0 High N1
N1 20 5 sports 1 1 High N1
23 0 sports 2 2 High N1
32 4 family 3 3 Low N1
43 2 truck 4
68 3
4 Low N1
family 5
5 High N1
 Evaluating Split Points

 Gini Index
◼ if data D contains examples from c classes

Gini(D) = 1 -  pj2
where pj is the relative frequency of class j in D
If D split into D1 & D2 with n1 & n2 tuples each
Ginisplit(D) = n1* gini(D1) + n2* gini(D2)
n n
Note: Only class frequencies are needed to compute index
 Finding Split Points

 For each attribute A do

◼ evaluate splits on attribute A using attribute list

 Key idea: To evaluate a split on numerical attributes we need

to sort the set at each node. But, if we have all attributes pre-
sorted we don’t need to do that at the tree construction phase

 Keep split with lowest GINI index

 Finding Split Points: Continuous Attrib.
 Consider splits of form: value(A) < x
◼ Example: Age < 17
 Evaluate this split-form for every value in an attribute list
 To evaluate splits on attribute A for a given tree-node:

Initialize class-histograms of left and right children;

for each record in the attribute list do

find the corresponding entry in Class List and the class and Leaf node
evaluate splitting index for value(A) < record.value;
update the class histogram in the leaf
 N1 GINI Index:
High Low
L 0 0
0 und
Class Leaf R 4 2
0
Age CLI
0 High N1
17 1
1 1 High N1
20 5 High Low
2 High N1
3
23 0
1
L 1 0 0.33
32 4 3 Low N1
4 R 3 2
43 2 4 Low N1
68 3 5 High N1
High Low
L 3 0
3
R 1 2
0.22
1: Age < 20

3: Age < 32 Age < 32

High Low
4: Age < 43
4 L 3 1
0.5
R 1 1
 Finding Split Points: Categorical Attrib.

 Consider splits of the form: value(A)  {x1, x2, ..., xn}

◼ Example: CarType  {family, sports}
 Evaluate this split-form for subsets of domain(A)
 To evaluate splits on attribute A for a given tree node:

initialize class/value matrix of node to zeroes;

for each record in the attribute list do

increment appropriate count in matrix;

evaluate splitting index for various subsets using the constructed matrix;
 class/value matrix

Car Type CLI Risk Leaf High Low

family 0 High N1 family 2 1
sports 1 High N1 sports 2 0
sports 2 High N1 truck 0 1
family 3 Low N1
truck 4 Low N1
family 5 High N1
Left Child Right Child GINI Index:

High Low High Low

CarType in {family} 2 1 2 1
GINI = 0.444

High Low High Low

CarType in {sports} 2 0 2 2
GINI = 0.333

High Low High Low

CarType in {truck} GINI = 0.267
0 1 4 1
 Updating the Class List

◼ Next step is to update the Class List with the new nodes
◼ Scan the attr list that is used to split and update the corresponding
leaf entry in the Class List

For each attribute A in a split traverse the attribute list

for each value u in the attr list
find the corresponding entry in the class list (e)
find the new node c to which u belongs
update node reference in e to the node corresponding to c
 Preventing overfitting

 A tree T overfits if there is another tree T’ that gives

higher error on the training data yet gives lower error
on unseen data.
 An overfitted tree does not generalize to unseen
instances.
 Happens when data contains noise or irrelevant
attributes and training size is small.
 Overfitting can reduce accuracy drastically:
◼ 10-25% as reported in Minger’s 1989 Machine
learning
 Approaches to prevent overfitting

 Two Approaches:
◼ Stop growing the tree beyond a certain point
◼ First over-fit, then post prune. (More widely used)
 Tree building divided into phases:

▪ Growth phase
▪ Prune phase
 Hard to decide when to stop growing the tree, so
second approach more widely used.
 Criteria for finding correct final tree size:

 Three criteria:
◼ Cross validation with separate test data
◼ Use some criteria function to choose best size
 Example: Minimum description length (MDL)
criteria
◼ Statistical bounds: use all data for training but apply
statistical test to decide right size.
 Occam’s Razor
 Given two models of similar generalization errors, one
should prefer the simpler model over the more complex
model
 Therefore, one should include model complexity when
evaluating a model

“entia non sunt multiplicanda praeter ecessitatem,”

which translates to:

“entities should not be multiplied beyond necessity.”

Minimum Description Length
(MDL) Yes
A?
No
X y 0 B?
X y
X1 1 B1 B2
X1 ?
X2 0 C? 1
A C1 C2 B X2 ?
X3 0
0 1 X3 ?
X4 1
X4 ?
… …
Xn
… …
1
Xn ?
 Cost(Model,Data) = Cost(Data|Model) +
Cost(Model)
◼ Cost is the number of bits needed for encoding.
◼ Search for the least costly model.
 Cost(Data|Model) encodes the misclassification
errors.
 Cost(Model) uses node encoding (number of
children) plus splitting condition encoding.
 Encoding data
 Assume t records of training data D
 First send tree m using L(m) bits
 Assume all but the class labels of training data
known.
 Goal: transmit class labels using L(D|m)
 If tree correctly predicts an instance, 0 bits
 Otherwise, log k bits where k is number of
classes.
 Thus, if e errors on training data: total cost
 e log k + L(m|M) bits.
 Complex tree will have higher L(m) but lower e.
 Question: how to encode the tree?
 SPRINT
 An improvement over SLIQ
 Does not need to keep a list in main memory
 Attribute lists are extended with class field – no Class list is
needed
 After a split, the ALs are partitioned.
 To split the ALs of the non-split attributes a hash table with
the record groups is kept in memory
 Pros and Cons of decision trees

• Pros • Cons
+ Reasonable training time – Cannot handle complicated
+ Fast application relationship between features
+ Easy to interpret – simple decision boundaries
+ Easy to implement – problems with lots of missing
+ Can handle large number data
of features
Decision Boundary
1

0.9

0.8
x < 0.43?

0.7
Yes No
0.6
y

0.5 y < 0.47? y < 0.33?

0.4

0.3
Yes No Yes No

0.2
:4 :0 :0 :4
0.1 :0 :4 :3 :0
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

x
• Border line between two neighboring regions of different classes is
known as decision boundary
• Decision boundary is parallel to axes because test condition involves
a single attribute at-a-time
Oblique Decision Trees

x+y<1

Class = + Class =

• Test condition may involve multiple attributes

• More expressive representation
• Finding optimal test condition is computationally expensive