Use of Distribution Algorithms, for the Construction of a Classification and Regression Tree

• Author: Adem Meta
• Position: Cuyahoga Community College, USA
• Pages: 45-51

Vo

l. 4 N

o

. 2

Jul

y

, 201

8

ISS

N 2410-391

8

A

cces online at www.ii

p

ccl.or

g

45

A

cademic Journal o

f

Business,

A

dministration, Law and Social Sciences

II

PCCL Publishin

g

, Graz-

A

ustria

Use of Distribution Algorithms, for the Construction of a Classifi cation and

Regression Tree

M

Sc.

Ade

m M

e

t

a

Cu

y

ahoga Communit

y

College,

U

S

A

Abstract

One of the most important processes in the construction of the classifi cation and regression

trees is the distribution of a given data. There are numerous algorithms for predicting

continuous variables or categorical variables from a set of continuous predictors and/or

categorical factor e ects.

I

n this paper

I

address the problem of learning various types o

f

algorithms to be used to get a optimal decision trees from data base.

I

n particular, we study

online machine learnin

g

al

g

orithms for learnin

g

classifi cation and re

g

ression trees, linear

model trees, option trees for regression, multi-target model trees, and ensembles of model

trees from a

g

iven data. Decision tree builds classifi cation or re

g

ression models in the form of a

tr

ee

str

u

ct

u

r

e

.

I

t br

ea

ks d

o

wn

a

d

a

t

a

s

e

t

i

nt

o

sm

a

ll

e

r

a

nd sm

a

ll

e

r s

u

bs

e

ts wh

i

l

e

a

t th

e

s

a

m

e

t

i

m

e

an associated decision tree is incrementally developed. The fi nal result is a tree with decision

n

odes and leaf nodes.

A

decision node has two or more branches. Leaf node represents a

c

lassifi cation or decision. The topmost decision node in a tree which corresponds to the best

p

redictor called root node. Decision trees can handle both categorical and numerical data. The

c

ore algorithm for building decision trees called, greedy search through the space of possible

b

r

a

nch

es

uses

E

ntrop

y

a

nd

I

n

f

ormation Gai

n

t

o

c

o

nstr

u

ct

a

d

e

c

i

s

io

n tr

ee

.In this paper, through

a concrete example, I will explicitly look at the use of four algorithms such as the Gini Index,

Chi-Square, Entropy and the Variance Reduction on which node will be the distribution of a

database. Once the data base is small, no doubt that the calculation is much more simple than

in the case of a large database.

Ke

y

words: distribution al

g

orithms, re

g

ression tree, classifi cation

.

Introduction

T

h

e

b

asi

c classifi cation and regression algorithms are considered to be one of the

best learning methods and used the most. Methods based on the classifi cation tree

provide predictive models with very good precision, stability and very ease of

interpretation. They represent non-linear links quite well and are suitable for solving

any classifi cation or regression problems. Decision trees use multiple algorithms to

decide when to split a node into two or more sub-nodes. The creation of subunits

increases the homogeneity of the resulting subunits. Thus, the purity of the node

increases with respect to the target variable. The crucial tree divides the nodes into all

available variables and then selects the resulting partition with the most homogeneous

sub nodes.

The choice of algorithms is also based on the type of responsible variables. Let’s look

at the four most used algorithms in the decision tree using the following example:

Let’s take a class of 36 students with three variables Gender (male/female), Class (XI

/ XII) and Height (160 cm up to 180 cm, (160,170) and (170,180)), 18 of which play

basketball at leisure. We want to create a model to predict who will play basketball