Classification Type Of Animal Based On Characteristic Using Weka

foulchilianΤεχνίτη Νοημοσύνη και Ρομποτική

20 Οκτ 2013 (πριν από 3 χρόνια και 8 μήνες)

257 εμφανίσεις

Classification Type Of Animal Based On
Characteristic Using Weka

Panca Kusuma Yuliandari.
#
1
,
Fitriannisa Umami.
#
2
,
Nur Fatih
#
3

#
Department of Information System
,
Institute Technology of Sepuluh Nopember

Surabaya, Indonesia

1
panca_kusuma_y

@

yahoo.com
,
2
neez_93
@

is.its.ac.id,
3
nur_fatih
@

is.its.ac.id



Abstract
-

It’s difficult to c
lassified animal according
spec
ies because there are many kind of classification.

A
nimal studies are classified into the type of mammal,
bird, reptile, fish,

amphibian,
insect, invertebrate
.

Variables

which used in this classification are feathers,
egg, milk, airbone, aquatic, predator, toothed,
backbone, breathes, Venomous, fins, legs, tail, domestic.

We use data from zoo.arff. There are 101 data consist of
training dat
a and testing data. In this study, the
classification performed using Backpropagation
artificial neural network software (software) Weka
(Waikato Environment Knowledge Analysis) 3.5.7. The
best results are obtained with the parameters of hidden
layer neuro
ns 16; learning rate 0.2; momentum of 0.2
and a maximum of 5000 iterations was 96.0396%.


Keyword

-

Animal,Artificial neural network,
weak

I.

INTRODUCTION


Artificial neural network is one branch of
artificial intelligence (artificial intelligence) that
is one of the information processing system
designed to mimic how the human brain to solve
a problem. In animal studies are classified into
the type of mammal
, bird, reptile, fish,
amphibian, insect, invertebrate data based on the
investigation. Artificial neural networks (ANN) is
a method that can be used to identify patterns in
the classification of animals into a species that are
mammal, bird, reptile, fish,

amphibian, insect or
invertebrate.The authors use the method of
Artificial Neural Networks with Backpropagation
software (software) Weka 3.5.7.

This study aims to develop models to classify the
types of animals based on the characteristic
possessed. The
model created using Neural
Network with Backpropagation algorithm.



II.

D
ISCUSION


Visualize all



Figure
1
. visualization

A.

LITERATURE REVIEW


1.

Metode Backpropagation


Backpropagation

is a supervised learning
algorithm and it is usually used by the
perceptron with many layers to change the
weights connected to neurons on hidden
layers. In general backpropagation model
is:


n

zj = f(v0j + Σ xi vij) ..............................
(1)

i=1

p

yk = f(w0k + Σ zj wjk) .............................
(2)

i=1




Description: (1)
activation function of the
input signals weighted xi
sent to the hidden layer.


(2)
activation function of
the sum of the input signals
weighted zj

sent to the
output

layer


2.

Network Architecture

Neural network consists of 3 layers: the
layer of input / inputs consist of 1 unit of
input variable nerve cells, a hidden layer
consisting of 16 units of nerve cells, and
the output layer and output con
sists of 7
neural cells. Input layer (xi) is used to
accommodate the variable that is 1
animal, while the output layer 7 (yk) is
used to present the classification for the
type of animal mammal, bird, reptile, fish,
amphibian, insect, invertebrate


Descrip
tion:


x = input (input)

j = 1 to n (n = 10).

v = Weight of the hidden layer.

w = weight of output layer.

n = number of processing units in the
hidden layer.

b = bias in the hidden layer and output
layer.

k = number of processing units in the
output

layer.

Y = Output result of mammal, bird,
reptile, fish, amphibian, insect,
invertebrate.


Backpropagation algorithm using the
output error to change the value
-
weight
weight in the back (backward).

1. Starting with the input layer, calculate
the output
of each processing element
through the outer layer.

2. Calculate the error in the outer layer
which is the difference between actual and
target data.

3. Transform the error in the
corresponding error in the input side of
the processing elements.

4. Make

these errors at the output of each
processing element to the error contained
in the input. This process repeats until the
input is reached.

5. Changing the weight by using the errors
on the input side of the superficial
elements and processing elements
are
connected.


3.

Animal Clasification

Classification system is based on the type
of animal species by looking at the
features that animal have.They will be
grouped by whether or not there hear,
feathers, egg, milk, airbone
, aquatic,
predator, toothed, backbone, breathes,
Venomous, fins, tail, domestic even how
many leg they have.According amount
characteristic of each animal owned by
the new type of animal can be classified in
to outline animal.They classified into
several
types, namely:


1.

Mammal

2.


Bird

3.

Reptile

4.

Fish

5.

Amphibian

6.


Insect

7.

invertebrata.



B.

RESEARCH METHODOLOGY


1.

Research Procedure

The data used in this study consisted of
101 data,which are training data and
testing data collected in an arff file. Weka

software there are three preprocessing
load, analyze, and filter. Load process, ,
data has been calculated using the formula
stored in to arff file. Zoo data is inserted
into the Weka open file panel. Analyze
the process, is used for variables analysis
an
imal. This animal classification process
does not use filters for preprocessing the
data is good for processing. Next to
classify the tab using the backpropagation
algorithm multilayer perceptron. In this
classify tab parameters used can be
modified.


To
analyze the backpropagation neural
network algorithm is analysze weight.
Each unit of input (xi, i = 1) receives the
signal xi and deliver signals to the hidden
layer units. Input units (xi) is Animal. Vij
is the weight multiplied by the unit value
of inp
uts (xi). Each hidden unit (xi, i =
1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16)
summed with the weight of the input
signal. v0j is bias in the hidden units.
Activation function to calculate the output
signal and sent to all units in the output
layer units. Eac
h unit of output (yk, k =
1,2,3,4,5,6,7) summed weighted input
signals. w0k is biased output unit.
Activation function to calculate the output
signal. Determine the output value mamal
animal, bird, reptile, fish, amphibian,
insect or invertebrate.


2.

Resear
ch Technique

Zoo data used is in arff format. parameters
that used to classify the animal type
consists of hidden layer neurons
1,2,3,4,5.; learning rate 0.01; 0.03; 0.05;
0.1; 0 .2. Momentum 0.2; and the number
of iteration 5000. Activation function use
d
is function Percepteron.Proses Multilayer
neural network with training
MultilayerPerceptron use test mode
options that crossvalidation training data
and testing data is processed into 10 parts
(folds) repeatedly. From the 101 data,
Weka software will sha
re and process data
9 / 10 for training data and 1 / 10 for
testing data. The process stops until the
fold to reach 10.



Figure
2
. classifier output


Figure
3
. output weight



Figure
4
. output weight


C.

Output

Classification

Results of soil classification with artificial
neural network method in the Weka
backpropagation

can be correctly classified
according to the table, the kappa statistic and
the confusion matrix.


Tabel 1.
Correctly Classified


LR/
HL

0,01

0,03

0,05

0,1

0,2

2

60.39
6 %

60.39
6 %

74.25
74 %

79.20
79 %

86.13
86 %

4

69.30
69 %

74.25
74 %

87.12
87 %

89.10
89 %

92.07
92 %

6

73.26
73 %

88.11
88 %

89.10
89 %

94.05
94 %

94.05
94 %

8

82.17
82 %

90.09
9 %

94.05
94 %

96.03
96 %


96.03
96 %

10

84.15
84 %

90.09
9 %

94.05
94 %

95.04
95 %

96.03
96 %


When tested using the Weka

software, proved
by the number of 16 hidden layer neurons and
learning rate of 0.2 obtained the largest
percentage is 96.0396%. So, from the 101
observed data there are 96 unclassified data is
correct, while the remaining 5 data in error
The second classi
fication consideration.
As a
second consideration would be seen much
kappa statistic of experiments that can be seen
in Table 2.







Tabel 2. Kappa Statistic


LR/H
L

0,01

0,03

0,05

0,1

0,2

2

0.44
81

0.41
98

0.63
86

0.71
5

0.81
47

4

0.57
75

0.64
68

0.82
57

0.85
42

0.89
53

6

0.63
82

0.84
19

0.85
56

0.92
16

0.92
16

8

0.76
25

0.86
74

0.92
15

0.94
76

0.94
77

10

0.78
76

0.86
83

0.92
15

0.93
46

0.94
77


In the table 7 can be seen that the number of
neurons 16 and 0.2 learning rate neural
networks can reach 0.9477. Value of kappa
statistic has a value between
-
1 and +1. Kappa
statistic value> = 0.75 indicates a very decent
result. 0 <kappa <0.4 shows reaso
nable results
for some levels. Kappa = 0, showed the same
results with random probability.

Kappa <0, indicates the probability of a
random bad and the result has no meaning. So,
with a kappa value of 0.9477 indicates
statistical achieve very decent result
s in the
truth of the classification. As a third
consideration would see confusion matrix

of the research experiment in table 3.



























Tabel 3. Confusion Matrix


LR/H
L

0,01

0,03

0,05

0,1

0,2

2


a b c d e f g <
--

classified as


41 0 0 0 0 0 0 | a
= mammal


0 20 0 0 0 0 0 | b
= bird


0 5 0 0 0 0 0 | c
= reptile


0 13 0 0 0 0 0 | d
= fish


0 4 0 0 0 0 0 | e
= amphibian


0 8 0 0 0 0 0 | f
= insect


0 10 0 0 0 0 0 |

g
= invertebrate


a b c d e f g <
-
-

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


3 2 0 0 0 0 0 |
c = reptile


4 9 0 0 0 0 0 |
d = fish


2 2 0 0 0 0 0 |
e = amphibian


5 3 0 0 0

0 0 |
f = insect


3 7 0 0 0 0 0 |
g = invertebrate

a b c d e f g <
-
-

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


1 2 0 1 0 0 1 |
c = reptile


0 0 0 13 0 0 0 |
d = fish


2 2 0 0

0 0 0 |
e = amphibian


5 2 0 0 0 0 1 |
f = insect


2 3 0 4 0 0 1 |
g = invertebrate

a b c d e f g <
-
-

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


1 1 0 3 0 0 0 |
c = reptile


0

0 0 13 0 0 0 |
d = fish


2 1 0 0 0 0 1 |
e = amphibian


2 2 0 0 0 2 2 |
f = insect


1 1 0 1 0 3 4 |
g = invertebrate

a b c d e f g <
-
-

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


0 1 0 1 0 0 3 |
c = reptile


0 0 0 13 0 0 0 |
d = fish


1 1 0 1 0 1 0 |
e = amphibian


0 0 0 0 0 7 1 |
f = insect


0 1 0 1 0 2 6 |
g = invertebrate

4


a b c d e f g <
--

classified as


41 0 0 0 0 0 0

| a
= mammal


0 20 0 0 0 0 0 | b
= bird


1 4 0 0 0 0 0 | c
= reptile


0 4 0 9 0 0 0 | d
= fish


0 0 0 4 0 0 0 | e
= amphibian


0 7 0 1 0 0 0 | f
= insect


1 4 0 5 0 0 0 | g
= invertebrate

a b c d e f g

<
-
-

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


0 1 0 4 0 0 0 |
c = reptile


0 0 0 13 0 0 0 |
d = fish


2 0 0 2 0 0 0 |
e = amphibian


2 3 0 3 0 0 0 |
f = insect


1 0 0 8 0 0

1 |
g = invertebrate

a b c d e f g <
-
-

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


1 1 0 3 0 0 0 |
c = reptile


0 0 0 13 0 0 0 |
d = fish


2 0 0 1 0 0 1 |
e = amphibian


0 1 0 0

0 5 2 |
f = insect


1 0 0 0 0 0 9 |
g = invertebrate


a b c d e f g
<
--

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


0 1 1 3 0 0 0 |
c = reptile


0 0 0 13 0 0 0 |
d = fish


2 0

1 1 0 0 0 |
e = amphibian


0 0 0 0 0 6 2 |
f = insect


0 0 0 0 0 1 9 |
g = invertebrate

a b c d e f g <
-
-

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


0 1 2 1 1 0 0 |
c = reptile


0 0 0 13 0 0 0 |
d = fish


0 0 2 0 1 1 0 |
e = amphibian


0 0 0 0 0 7 1 |
f = insect


0 0 0 0 0 1 9 |
g = invertebrate

6


a b c d e f g <
--

classified as


41 0 0 0 0 0 0 | a
= mammal


0 20 0 0 0 0 0

| b
= bird


0 1 0 4 0 0 0 | c
= reptile


0 0 0 13 0 0 0 | d
= fish


0 0 0 4 0 0 0 | e
= amphibian


0 4 0 4 0 0 0 | f
= insect


1 0 0 9 0 0 0 | g
= invertebrate

a b c d e f g <
-
-

classified as


41 0 0 0 0

0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


0 3 0 2 0 0 0 |
c = reptile


0 0 0 13 0 0 0 |
d = fish


0 0 1 2 0 0 1 |
e = amphibian


0 0 0 0 0 8 0 |
f = insect


0 0 0 0 0 3 7 |
g = invertebrate

a b c d e

f g <
-
-

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


0 1 0 3 1 0 0 |
c = reptile


0 0 0 13 0 0 0 |
d = fish


0 0 2 2 0 0 0 |
e = amphibian


0 0 0 0 0 8 0 |
f = insect


0 0 0 0

0 2 8 |
g = invertebrate


a b c d e f g <
-
-

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


0 1 3 1 0 0 0 |
c = reptile


0 0 0 13 0 0 0 |
d = fish


0 0 2 0 2 0 0 |
e = amphibian


0 0

0 0 0 8 0 |
f = insect


0 0 0 0 0 2 8 |
g = invertebrate

a b c d e f g <
-
-

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


0 1 2 1 1 0 0 |
c = reptile


0 0 0 13 0 0 0 |
d = fish


0

0 1 0 3 0 0 |
e = amphibian


0 0 0 0 0 8 0 |
f = insect


0 0 0 0 0 2 8 |
g = invertebrate

8

a b c d e f g <
--

classified as


41 0 0 0 0 0 0 | a
= mammal


0 20 0 0 0 0 0 | b
= bird


0 2 0 2 0 0

1 | c
= reptile


0 0 0 13 0 0 0 | d
= fish


0 0 0 1 0 0 3 | e
= amphibian


0 0 0 0 0 0 8 | f
= insect


0 0 0 0 0 1 9 | g
= invertebrate

a b c d e f g <
-
-

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0

0 0 0 0 |
b = bird


0 2 0 2 1 0 0 |
c = reptile


0 0 0 13 0 0 0 |
d = fish


2 0 1 1 0 0 0 |
e = amphibian


0 0 0 0 0 8 0 |
f = insect


0 0 0 0 0 1 9 |
g = invertebrate


a b c d e f g
<
--

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


0 1 3 1 0 0 0 |
c = reptile


0 0 0 13 0 0 0 |
d = fish


0 0 3 0 1 0 0 |
e = amphibian


0 0 0 0 0 8 0 |
f = insect


0 0 0 0 0 1 9 |
g = invertebrat
e


a b c d e f g <
-
-

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


0 1 3 1 0 0 0 |
c = reptile


0 0 0 13 0 0 0 |
d = fish


0 0 1 0 3 0 0 |
e = amphibian


0 0 0 0 0 8 0 |
f = inse
ct


0 1 0 0 0 0 9 |
g = invertebrate

a b c d e f g <
-
-

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


0 1 3 1 0 0 0 |
c = reptile


0 0 0 13 0 0 0 |
d = fish


0 0 1 0 3 0 0 |
e = a
mphibian


0 0 0 0 0 8 0 |
f = insect


0 0 0 0 0 1 9 |
g = invertebrate

10

a b c d e f g <
--

classified as


41 0 0 0 0 0 0 | a
= mammal


0 20 0 0 0 0 0 | b
= bird


1 1 0 2 0 1 0 | c
= reptile


a b c d e f g
<
--

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


0 1 0
2 1 0 1 |
c = reptile


a b c d e f g
<
--

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


0 1 3 1 0 0 0 |
c = reptile

a b c d e f g <
-
-

classifi
ed as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


0 1 3 1 0 0 0 |
c = reptile


a b c d e f g
<
--

classified as


41 0 0 0 0 0 0 |
a = mammal


0 20 0 0 0 0 0 |
b = bird


0 1 3 1 0 0 0 |
c = reptile


In table 3 can be seen that the number of
hidden layer neurons, learning rate 16 and
confusion matrix 0.2 has a very feasible to
see the distribution of the error classification.
In the
confusion matrix, which determines the
amount of truth classification is indicated by
the formation of the diagonal of the matrix.


REFERENCES


Nafisah, Sari, Puspitodjati Sulistyo, and
Wulandari Sri,

“pengklasifikasian jenis tanah
menggunakan jaringan
saraf tiru
an dengan
menggunakan algoritma
backpropagation”,2008.




0 0 0 13 0 0

0 | d
= fish


0 0 0 1 0 2 1 | e
= amphibian


0 0 0 0 0 2 6 | f
= insect


1 0 0 0 0 0 9 | g
= invertebrate


0 0 0 13 0 0 0 |
d = fish


1 0 2 0 0 1 0 |
e = amphibian


0 0 0 0 0 8 0 |
f = insect


0 0 0 0 0 1 9 |
g = invertebrate


0 0 0 13 0 0 0 |
d = fish


0 0 3 0 1 0 0 |
e = amphibian


0 0 0 0 0 8 0 |
f = insect


0 0 0 0 0 1 9 |
g = invertebrate


0 0 0 13 0 0 0 |
d = fish


0 0 1 0 3 0 0 |
e = amphibian


0 0 0 0 0 7 1 |
f = insect


0 0 0 0 0 1 9 |
g = inver
tebrate


0 0 0 13 0 0 0 |
d = fish


0 0 1 0 3 0 0 |
e = amphibian


0 0 0 0 0 8 0 |
f

= insect


0 0 0 0 0 1 9 |
g = invertebrate