Induction of Decision Trees

Peter Waiganjo Wagacha

This notes are for ICS320 Foundations of Learning and Adaptive Systems

Institute of Computer Science

University of Nairobi
PO Box 30197, 00200 Nairobi.

9th May, 2003

1
Contents
1 Induction of Decision Trees 3

2 Introduction 3

3 What is a decision tree? 3

4 Constructing decision trees 4

5 Which attribute is the best classifier? 6

6 Entropy 6

7 Information gain 8

8 Incorporating Continuous-valued attributes 9

9 How Decision trees partition the space 9

10 Alternative measures for selecting attributes 9

10.1 Gain ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
10.2 Gini index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

11 Strengths and Weaknesses of Tree methods 13

11.1 Strengths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
11.2 Weaknesses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2
1 Induction of Decision Trees
2 Introduction
Decision trees are powerful and popular tools for classification and prediction.
Decision trees are attractive due to the fact that, in contrast to other machine
learning techniques such as neural networks, they represent rules. Rules can
readily be expressed so that humans can understand them or even directly used
in a database access language like SQL so that records falling into a particular
category may be retrieved.

In some applications, the accuracy of a classification or prediction is the only

thing that matters. In such situations we do not necessarily care how or why
the model works. In other situations, the ability to explain the reason for a
decision, is crucial. In medical diagnosis tests are regularly performed. In many
cases, there is often more than one test for a particular series of tests. Often one
test may be more accurate than another. One test may also be more expensive
than another. To the patient, all these may be irrelevant, but the doctor has to
bear these intricacies in mind when interpreting the test results. In marketing
one has to describe the customer segments to marketing professionals, so that
they can utilize this knowledge in launching a successful marketing campaign.
These domain experts must recognize and approve this discovered knowledge,
and for this we need good descriptions. There are a variety of algorithms for
building decision trees that share the desirable quality of interpretability. A well
known and frequently used over the years is C4.5 developed by J. Ross Quinlan
[Qui93] (or improved, but commercial version See5/C5.0).

3 What is a decision tree?

Decision tree is a classifier in the form of a tree structure (see Figure 1), where
each node is either:

• a leaf node – indicates the value of the target attribute (class) of exam-
ples, or

• a decision node – specifies some test to be carried out on a single

attribute-value, with one branch and sub-tree for each possible outcome
of the test.

A decision tree can be used to classify an example by starting at the root of the
tree and moving through it until a leaf node, which provides the classification
of the instance.

Decision tree induction is a typical inductive approach to learn knowledge on

classification. The key requirements to do mining with decision trees are:

3
 Figure 1: Decision tree: for conditions to play tennis.

• Attribute-value description: object or case must be expressible in

terms of a fixed collection of properties or attributes. This means that we
need to discretize continuous attributes, or this must have been provided
in the algorithm.

• Predefined classes (target attribute values): The categories to which

examples are to be assigned must have been established beforehand (su-
pervised data).

• Discrete classes: A case does or does not belong to a particular class,

and there must be more cases than classes.

• Sufficient data: Usually hundreds or even thousands of training cases.

4 Constructing decision trees

Most algorithms that have been developed for learning decision trees are varia-
tions on a core algorithm that employs a top-down, greedy search through the
space of possible decision trees. Decision tree programs construct a decision tree
T from a set of training cases.

J. Ross Quinlan originally developed ID3 at the University of Sydney. He first

presented ID3 in 1975 in a book, Machine Learning, vol. 1, no. 1. ID3 is based
on the Concept Learning System (CLS) algorithm.

4
Function ID3
Input:
(R: a set of non-target attributes,
C: the target attribute,
S: a training set) returns a decision tree;

begin
If S is empty, return a single node with value Failure;
If S consists of records all with the same value for the
target attribute, return a single leaf node with that value;
If R is empty, then return a single node with the value of
the most frequent of the values of the target attribute that are found
in records of S; [in that case there may be be errors, examples
that will be improperly classified];
Let A be the attribute with largest
Gain(A,S) among attributes in R;
Let {aj | j = 1, 2, · · · , m} be the values of attribute A;
Let {Sj | j = 1, 2, · · · , m} be the subsets of S consisting
respectively of records with value aj for A;
Return a tree with root labelled A and arcs labelled a1 , a2 , · · ·, am
going respectively to the trees (ID3 (R-{A}, C, S1 ),
ID3 (R-{A}, C, S2 ), · · ·, ID3 (R- {A}, C, Sm );
Recursively apply ID3 to subsets Sj | j = 1, 2, · · · , m until they are empty
end

Table 1: ID3 Decision Tree Algorithm

5
ID3 searches through the attributes of the training instances and extracts the
attribute that best separates the given examples. If the attribute perfectly clas-
sifies the training sets then ID3 stops; otherwise it recursively operates on the
m (where m = number of possible values of an attribute) partitioned subsets to
get their “best” attribute. The algorithm uses a greedy search, that is, it picks
the best attribute and never looks back to reconsider earlier choices. Note that
ID3 may misclassify data.

The central focus of the decision tree growing algorithm is selecting which at-
tribute to test at each node in the tree. For the selection of the attribute with
the most inhomogeneous class distribution the algorithm uses the concept of
entropy, which is explained next.

5 Which attribute is the best classifier?

The estimation criterion in the decision tree algorithm is the selection of an
attribute to test at each decision node in the tree. The goal is to select the
attribute that is most useful for classifying examples. A good quantitative
measure of the worth of an attribute is a statistical property called information
gain that measures how well a given attribute separates the training examples
according to their target classification. This measure is used to select among
the candidate attributes at each step while growing the tree.

6 Entropy
In order to define information gain precisely, we need to define a measure
commonly used in information theory, called entropy, that characterizes the
(im)purity of an arbitrary collection of examples. Given a set S, containing
only positive and negative examples of some target concept (for a 2 class prob-
lem), the entropy of set S relative to this simple, binary classification is defined
as:

Entropy(S) = −pp log2 pp − pn log2 pn (1)

where pp is the proportion of positive examples in S and pn is the proportion

of negative examples in S. In all calculations involving entropy we define 0log0
to be 0.

To illustrate, suppose S is a collection of 25 examples, including 15 positive

and 10 negative examples [15+, 10-]. Then the entropy of S relative to this
classification is

15 15 10 10
Entropy(S) = −( )log2 ( ) − ( )log2 ( ) = 0.970 (2)
25 25 25 25

6
Notice that the entropy is 0 if all members of S belong to the same class. For
example, if all members are positive (pp = 1), then pn is 0, and Entropy(S) =
-1∗log2 (1) - 0∗log2 0 = -1∗0 - 0∗log2 0 = 0. Note the entropy is 1 (at its maxi-
mum!) when the collection contains an equal number of positive and negative
examples. If the collection contains unequal numbers of positive and nega-
tive examples, the entropy is between 0 and 1. Figure 2 shows the form of the
entropy function relative to a binary classification, as p+ varies between 0 and 1.

1.0
Entropy(S)

0.5

0.0 0.5 1.0

p
Figure 2: The entropy function relative to a binary classification, as the pro-
portion of positive examples pp varies between 0 and 1.

One interpretation of entropy from information theory is that it specifies the

minimum number of bits of information needed to encode the classification of
an arbitrary member of S (i.e., a member of S drawn at random with uniform
probability). For example, if pp is 1, the receiver knows the drawn example will
be positive, so no message need be sent, and the entropy is 0. On the other
hand, if pp is 0.5, one bit is required to indicate whether the drawn example is
positive or negative. If pp is 0.8, then a collection of messages can be encoded
using on average less than 1 bit per message by assigning shorter codes to col-
lections of positive examples and longer codes to less likely negative examples.

Thus far we have discussed entropy in the special case where the target classi-
fication is binary. If the target attribute takes on c different values, then the
entropy of S relative to this c-wise classification is defined as

7
 c


Entropy(S) = −pi log2 pi (3)
i=1

where pi is the proportion of S belonging to class i. Note the logarithm

is still base 2 because entropy is a measure of the expected encoding length
measured in bits. Note also that if the target attribute can take on c possible
values, the maximum possible entropy is log2 c.

7 Information gain
Given entropy as a measure of the impurity in a collection of training examples,
we can now define a measure of the effectiveness of an attribute in classifying the
training data. The measure we will use, called information gain, is simply the
expected reduction in entropy caused by partitioning the examples according to
this attribute. More precisely, the information gain, Gain(S, A) of an attribute
A, relative to a collection of examples S, is defined as

| Sv |
Gain(S, A) = Entropy(S) − Entropy (Sv ) (4)
|S|
v∈V (A)

where V alues(A) is the set of all possible values for attribute A, and Sv is the
subset of S for which attribute A has value v (i.e., Sv = {s S | A(s) = v}).
Note the first term in the equation for Gain is just the entropy of the original
collection S and the second term is the expected value of the entropy after S
is partitioned using attribute A. The expected entropy described by this sec-
ond term is simply the sum of the entropies of each subset Sv , weighted by the
fraction of examples | Sv |/| S | that belong to Sv . Gain(S, A) is therefore the
expected reduction in entropy caused by knowing the value of attribute A. Put
another way, Gain(S, A) is the information provided about the target attribute
value, given the value of some other attribute A. The value of Gain(S, A) is the
number of bits saved when encoding the target value of an arbitrary member of
S, by knowing the value of attribute A.

The process of selecting a new attribute and partitioning the training examples
is now repeated for each non-terminal descendant node, this time using only
the training examples associated with that node. Attributes that have been
incorporated higher in the tree are excluded, so that any given attribute can
appear at most once along any path through the tree. This process continues
for each new leaf node until either of two conditions is met:
1. every attribute has already been included along this path through the tree,
or
2. the training examples associated with this leaf node all have the same
target attribute value (i.e., their entropy is zero).

8
8 Incorporating Continuous-valued attributes
The initial definition of ID3 is restricted to attributes that take on a discrete
set of values. First, the target attribute whose value is predicted by the learned
tree must be discrete valued. Second, the attributes tested in the decision nodes
of the tree must also be discrete valued. This second restriction can easily
be removed so that continuous-valued decision attributes can be incorporated
into the learned tree. This can be accomplished by dynamically defining new
discrete-valued attributes that partition the continuous attribute value into a
discrete set of intervals. In particular, for an attribute A that is continuous-
valued, the algorithm can dynamically create a new Boolean attribute Ac that
is true if A < c and false otherwise. The only question is how to select the best
value for the threshold c. Clearly, we would like to pick a threshold, c, that
produces the greatest information gain. By sorting the examples according to
the continuous attribute A , then identifying adjacent examples that differ in
their target classification, we can generate a set of candidate thresholds midway
between the corresponding values of A. It can be shown that the value of c
that maximizes information gain must always lie at such a boundary. These
candidate thresholds can then be evaluated by computing the information gain
associated with each. The information gain can then be computed for each of
the candidate attributes, and the best can be selected. This dynamically created
Boolean attribute can then compete with the other discrete-valued candidate
attributes available for growing the decision tree.

9 How Decision trees partition the space

There are a couple of ways through which this is done. These methods revolve
around how to select attributes, which in turn dictates how the decision tree is
grown.

10 Alternative measures for selecting attributes

There is a natural bias in the information gain measure that favours attribute
with many values over those with few values. As an extreme example, consider
an attribute that perfectly predicts the target function over the training data,
that is; the test at the root creates as many links to a leaf as there are training
examples. Thus, each training example is represented by a single leaf, after
effecting the test at the root node of the tree. An example of such an attribute
can be the (unique) timestamp an event occurred.

The problem with this decision tree is that it will be a poor predictor on subse-
quent examples, despite the fact that it has a very high information gain over
the training data.

9
 y
1

x
< 0.8 > 0.8
0 x y
0 A 1
< 0.2 > 0.2
x
< 0.2 > 0.2
x
< 0.5 > 0.5
y y
< 0.5 > 0.5 < 0.6 > 0.6
B x
< 0.6 > 0.6
y
< 0.8 > 0.8

Figure 3: Decision tree partition

One way to avoid this difficulty is to select attributes based on other measure
than information gain [MKS94].

10.1 Gain ratio

The gain ratio measure penalizes attributes such as timestamp by incorporating
a term, called split information, that is sensitive to how broadly and uniformly
the attribute splits the data:
c

Si Si
SplitInf ormation(S, A) = − log2 (5)
i=1
S S
where S1 through Sc are the c subsets of examples resulting from partitioning
S by the c-valued attribute A. Note that SplitInf ormation is actually the en-
tropy of S with respect to the values of attribute A. This is in contrast to the
previous use of the entropy, in which the entropy of S was only considered with
respect to the target attribute whose value is to be predicted by the learned tree.

10
 x1
2.00
x1 > 1.25
A B
1.50
yes no

1.00
x2 > 1.00 x2 > 1.50
C D
yes no yes no 0.50

B A D C 0.00 x2
0.00 0.50 1.00 1.50 2.00

Figure 4: The left side of the figure shows a simple axis-parallel tree that uses
two attributes. The right side shows the partitioning that this tree creates in
the attribute space.

x1 x1

x2 x2

Figure 5: The left side shows a simple 2-D domain in which two oblique hyper-
planes define the classes. The right side shows an approximation of the sort
that an axis-parallel decision tree would have to create to model this domain.

11
The GainRation measure is defined in terms of the earlier Gain measure, as
well as this SplitInf ormation, as follows

Gain(S, A)
GainRatio(S, A) = (6)
SplitInf ormation(S, A)
Notice that the SplitInf ormation term discourages the selection of attributes
with many uniformly distributed values. For example, consider a collection of
n examples that are completely separated by attribute A. In this case, the
SplitInf ormation value will be log2 n. In contrast, a boolean attribute B that
splits the same n examples in half will have SplitInf ormation of 1. If attributes
A and B produce the same information gain, then clearly B will score higher
according to the GainRatio measure.

10.2 Gini index

In each of the following methods, the set of examples S at the node to be split
contains n instances that belong to one of k categories. Assume the example
set S is split into two non-overlapping subsets SL and SR . Lj and Rj are the
number of instances of category j in SL and SR respectively.

The GiniIndex was proposed for decision trees by Breiman. The Gini Index
as originally defined, measures the probability of mis-classification of a set of
instances, rather than the impurity of a split. Therefore, a little variation of the
original equation is provided:

k
Li 2
GiniL = 1 − ( ) (7)
i=1
|SL |

k
Ri 2
GiniR = 1 − ( ) (8)
i=1
|S R|

|SL | × GiniL + |SR | × GiniR

Impurity = (9)
n
where GiniL is the Gini Index for the subset SL and GiniR is that for the
subset SR .

Twoing Value The Twoing value was first proposed by Breiman. This value
is computed as follows

k
|SL | |SR | Li Ri 2
T woingV alue = × ×( | − |) (10)
n n i=1
|SL | |SR |
The T woingV alue is actually a goodness measure rather than an impurity mea-
sure. Therefore, the reciprocal of this value is often used as an indication of the
impurity.

12
11 Strengths and Weaknesses of Tree methods
Below we have enumerated the strengths and weaknesses of decision trees meth-
ods.

11.1 Strengths
The strengths of decision tree methods are:
• Decision trees are able to generate understandable rules.
• Decision trees perform classification without requiring much computation.
• Decision trees are able to handle both continuous and categorical variables.
• Decision trees provide a clear indication of which fields are most important
for prediction or classification.

11.2 Weaknesses
The weaknesses of decision tree methods are:
• Decision trees are less appropriate for estimation tasks where the goal is
to predict the value of a continuous attribute.
• Decision trees are prone to errors in classification problems with many
classes and a relatively small number of training examples.
• Decision tree can be computationally expensive to train. The process of
growing a decision tree is computationally expensive. At each node, each
candidate splitting field must be sorted before its best split can be found.
In some algorithms, combinations of fields are used and a search must
be made for optimal combining weights. Pruning algorithms can also be
expensive since many candidate sub-trees must be formed and compared.
• Decision trees do not treat non-rectangular classification regions well.
Most decision-tree algorithms only examine a single attribute (feature)
at a time. This leads to rectangular classification boxes that may not cor-
respond well with the actual distribution of records in the decision space.

References
[KBM96] Miroslav Kubat, Ivan Bratko, and Ryszard Michalski. A review of ma-
chine learning methods. Machine Laearning and Data Mining: Meth-
ods and Applications, 1996. Editors: R.S. Michalski and I. Bratko
and M. Kubat.
[MKS94] S. K. Murthy, S. Kasif, and S. Salzberg. A system for induction
of oblique decision trees. Journal of Artificial Intelligence Research,
2:1–32, 1994.

13
[Qui93] J. R. Quinlan. C4.5: Programs for Machine Learning. Morgan Kauf-
mann, San Mateo CA, 1993.

[RN95] Stuart Russell and Peter Norvig. Artificial Intelligence:

A Modern Approach. Prentice Hall, 1995. available at
http://www.aima.cs.berkeley.edu.

14