A Comparison of Backpropagation and LVQ · a case study of lung sound recognition

PROCEEDINGS

.

-

-

..

ｾ＠

"

.

-

-ICACSIS

2014

201 4 Intern ational Conference on

Advanced Computer Science and

lnforrnation Systems

October 18th and 19th 2014

Ambhara Hotel, Blok M

Jakarta, Indonesia

ISBN : 978· 979- 1421-22·5

I

)

;

i

1

-'

'

,.

I

I

I

f

l

l

ｾ ﾷ ｾ＠ ;•\ <.it;•;1.• ｊｴｴＱｩＭＮＮＺｩ｜ＮＮｴＱ｜ＱｾｵｵＬｵｴ･ａ＠

G

FE ·PUSTAKAAN

... r, II i ＭＮｾ＠ t•,.. "') 1 11 r PC.JC r A"'lA"" fU':c.c:>R

.... u: .. ·vs C• RW .. \v.& •' • r .. tp •f • • 101'' ) ' l!Ut4

f tn••' ﾵｷｾＱ ＰＮ｡Ｍｾ Ｑ ｩｬﾰＧＢＧ＠ .,..._..po« od uw." ttllpit ••w--• ﾷｴ ｾｯｭ＠ ' ""'P' •t1b ec id

+.IEEE

INDONESIA SECTION

ICACSIS

,•

2014 International Conference on

Advanced Computer Science and Information Systems

(Proceedings)

ISBN : 978-979 -1421-225

1

t

\

l

I

I

.

l

l

IC.\CSIS :!Ill I

plenary and invited spenkcrs.

Welcome Message from

General Chairs

On hc:half

,,r

the Organi1i11g Committee of this

ｬｮｴ｣ｲｮ｡ｴｩｾｲｮ｡ｬ＠

Conference on /\d\ anred Computer

Scicrl<.:c

anti

lnfonna1io11

S)stems

2014 (

IC/\CSIS

2014). \\C

\\lllllJ like

In

C:\lcnJ our \\ilrm ''ekomc

h.l

all uf 1hc pn.:sen1er anJ punicipants. and in particular.

\\C

\\OulJ like to express our sincere gratitude to our

This international con fcrence is organized

ｨｾ＠

the F acuh) of Computer Science, Univcrsitas

Indonesia. and is intentlcd to be the tirsl step to\i,,arJs a lop <:lass conference on Computer

Science and Information Systems. We hdieve that

ｴｨｩｾ＠

international conference "ill giH.:

opponunitics for sharing and exchanging original res\!arch ideas and opinions. gaining

inspiration for future research. and broadening knowledge about various fields in advanced

computer science and information s) stems. amongst members of lnJonesian research

communities. together \\ith researchers from Germany. Singapore. Thailand, France, Algeria.

Japan, Malaysia, Philippines. United Kingdom. S\\cdcn. United States and ot her countries.

This conference focuses on the tkvd opmcnl of computer science and informution

ｳｹｾ ｴ ｣ｭｳＮ＠

A long with 4 plenary and 2 invited speeches, the proceedings of this conference contains

71

papers which have been selected from a total of 132 papers from twelve different countries.

These selected papers will be presented during the cnnfercnce.

We also want to express our sincere appreciation to the members of the Program Comrniltec for

their critical review of the submitted papers, as \\ell as the Organizing Committee for the time

and energy they have devoted to editing the proceedings and arranging the logistics of holding

this conference. We would also like to give :ippreciation to the authors who have submitted their

excellent works to this conference. Last but not least. we would like to extend our gratitude to

the Ministry of Education of the Republic of Indonesia. the Rector of Un iversitas Indonesia.

Universitas Tarumanagara. l3ogor /\gricultural Institute. and the Dean of the Facult) of

Computer Science fo r their continued support towards the ICACSIS

1014

conference.

Sincerely yours.

General Chairs

IC 'AC 'SIS 21111

Welcorne Message from

The Dean of Faculty of Con1puter Science, Universitas

Indonesia

On

behalf of all the

academic

staff anJ

ｾｴｬｬｬｨＺｮｴｳ＠

of

the Faculty of Computer

Sc ience. Univcrsit:.i::. lnduncsia. I \\Ould like to exlc:nd our \\armest v.elcomc lo

all

the part icipants

w

the :\mbhara

I

lotd.

Jakarta

on the occasion of the

2014

International

C

onfcrem:c on Atlvam.:ctl Computer Sc icnce and Information

S)stems

(ICJ\CSIS).

Just like the previous fivc events in this series ( ICACSIS

2009. 2010, 2011. 2012,

anti

2013).

I

am

confident that

IC

/\S IS

201-t

v.

ill pla) an importanl rnle in t:ncouraging activities in n:search

and development of computer science and information technology in Indonesia. and give an

excellent opportunity to forge collaborations bcl\veen research institutions both \\ ithin the

country and with international partners. The

broad

scope

of

this event. \vhich includes both

theort:tical aspects of computer science anJ practical. applied experience

of

Jc\ eloping

information systems. provides a unique.: mccting ground for rcscarchcrs spanning tht: \\huk

spectrum of our discipline. I hope that

O\

er the next t\\O da)s. some: fruitful collaborations can

be

established .

I

also hope that the special attention devoted this year

to

the

lidd

nf pervasive computing,

including

the very

t>xciting

area of

wireless sensor networks. \viii ignite the development

uf

applications in th is area to address the various needs of

I

ndoncsia ·

s

development.

I

would like to express my sincere gratitude

lo

the distinguished invited speakers for thc.!ir

presence and contributions to the conferencc.

I

also thank all the program committee members

for their efforts in ensuring

a

rigorous review process to select high quality papers.

Finally.

I

sincerely hope that all the participants \\.ill benefit from the technical contents or this

conference. and wish you a very succcssful conference :-ind an enjoyable stay in Jakarta.

Sincere!)'.

Mirna Ad r ia ni, Ora, Ph.D.

Dean

of the

Faculty

of

Computer Science

Universitas Ind onesia

111

-

- -

- - -

-

t

I( '.\('SIS '211 l I

COMMITTEES

Honorary Chairs:

• A. Jain, I EE E Fellow, Mich igan State University,

us

T Fukuda , IEEE Fellow, Nagoya Un1vers1ty, JP

• M. Adnan1, Un1vers1tas Indonesia, JO

General Chairs:

E. K. Budiarjo, Un1versitas Indonesia, ID 0 . I Sensuse, Universitas Indonesia, ID Z. A. Has1buan, Universitas Indonesia, ID

Program Chairs:

H. B. Santoso, Un1versitas Indonesia, ID

• W. Jatm1ko, Unlversitas Indonesia, IO

• A. Buono, lnstitut Pertan1an Bogor, ID

0. E. Herwindiati, Universitas Tarumanegara, 10

Section Chairs:

K. Wastuwibowo, IEEE Indonesia Section, ID

Publication Chairs:

• A. Wibisono, Universitas Indonesia, ID

Program Committees:

A. Azurat, Universitas Indonesia, ID

p --·- - -·-- -

._ ..

Ｌｾ ＺＭＭﾷ Ｎ］Ｚ＠ - -- I

r

Ｎ ｾＧﾰ＠

,,

,;

c :· ｾＬ＠ ·:-; ｾｾ Ｎ ａＮｲＬｎ＠ ·

1 .. • ' \ t t • , • \\V •\ ltoe

)F..-.->A.,•,1t.01f,.' I. ot•:..•.1,·.•, ｾＬＨＧＢＢＬＨＩｒ＠

, .. _ . ...,/ Ｍ Ｎ［ＮＬｊｲ Ｎﾷ ｾｾｲＮ｣ＮｩＮ＠

-, -, ...... .:,. t t l ' l l : .. ｾｾ ｦ［ Ｎａ＠

E "'"' '' ' 11\ t •> A ... ＮＦｾ Ｇ ＧＭｃＮＺＺＢ＠ • • ｾ＠ ... tpb a.c •d •

',,.1-11 ... . , .. ﾷ ﾷＭｾ＠ •••• ' . ' • .. '

A. Fanar, ｌ･ｲｮ｢｡ｾ｡＠ llmu Pengetahuan Indonesia, ID

• A. Kist1Jantoro, Institut Teknolo91 Bandung, ID

A. Purwarianti, Institut Teknologi Bandung, ID

• A. Nugroho, PTJK BPPT, ID

• A. Srivihok, Kasetsart University, TH

• A. Arifin Institut Teknologi Sepuluh Nopember, ID

• A. M. Arymurthy, Unlversitas Indonesia, ID

A. N. H1dayanto, Universitas Indonesia, ID

• B. Wijaya, Univers1tas Indonesia, ID

• B. Yuwono, Un1versitas Indonesia, ID

B. Hard1an, Univers1tas Indonesia, ID

• B. Purwandari, Urnversitas Indonesia, ID

• B. A..,Nazief, Universitas Indonesia, ID

B. H. Widjaja, Universitas Indonesia, ID Denny, Urnversitas Indonesia, ID

• O. Jana, Computer Society of India, IN

• E. Gaura, Coventry University, UK

E. Seo, Sungkyunkwan University, KR F. Gaol, IEEE I ndonesia Section, ID

I('.\( ﾷＭＮＺ Ｑ ｾ＠ :2111 1

ICAC5/5 2014

f

•

H Manurung, Un1vers1tas Indonesia , ID

•

H. Suhartanto, Univers1tas Indones1d, ID

•

H. Sukoco, Jnst1tut Pertan1an Bogor, ID

•

H. Tolle, Univcrs1tas Braw11aya, ID

I Bud1, Un1vers1tas Indonesia, ID

•

I. S1tan9gang, lnst1tut Pertanian Bogor, 10

»

I Was1to, Univers1tas Indonesia , ID

•

K. Sekiyama, Nagoya Un1vers1ty, JP

•

L Stefanus, Univers1tas Indonesia, ID

Marim1n, Inst1tut Pertanian Bogor, ID

•

M T. Suarez, De La Salle University, PH

•

M. Fanany, Un1vers1tas Indonesia, I D

•

M. Kyas, Freie Un1vers1tat Berltn, DE

t'>

•

M. NakaJima, Nagoya University, JP

..

•

M. W1dyanto, Universitas Indonesia, ID

•

M. W1d1a1a, PTIK BPPT, ID

•

N. Maultdev1, Institut Teknolog1 Bandung, ID

•

0 . S1dek, Un1vers1t1 Sa1ns Malaysia, MY

0 Lawanto, Utah State University, US

•

P. Hitzler, Wnght State University, US

•

P. Mursanto, Universitas I ndonesia, ID ｾ＠

ｾ＠

•

S. Bressan, National University of Singapore, SG

•

S. Kuswad1, lnst1tut Teknolog1 Sepuluh Nopember, ID

•

S. Nomura, Nagaoka University of Technology, JP

•

S. Yaz1d, Universitas Indonesia, 10

•

T. Basaruddin, Un1vers1tas Indonesia, ID

•

T. Hardjono, Massachusetts Institute of Technology, US

T. Gunawan, International Islamic University Malaysia, MY

•

T . A. Masoem, Univers1tas Indonesia, ID

•

V. Allan, Utah State University, US

W. Chutimaskul, King Mokut's University of Technology, TH

•

W . Molnar, Public Research Center Henri Tudor, LU

•

W. Nugroho, Universitas Indonesia, ID

•

W . Prasetya, Universiteit Utrecht, NL

•

W. Sed1ono, International Islamic University Malaysia, MY

•

W. Sus1lo, University of Wollongong, AU

•

W. Wibowo, Universitas Indonesia, JD

•

X. Lt, The University of Queensland, AU

Y. lsal, Universitas Indonesia, ID

•

Y. Sucahyo, Un1vers1tas Indonesia, ID

Local Organ izing Committee:

•

A . Y. Utomo, Universitas Indonesia, ID

•

Aprinald1, Univers1tas Indonesia, ID

D. Eka, Un1vers1tas Indonesia, ID E. S. Utama , Universitas Indonesia, ID

•

E. Cahyanings1h, Universitas Indonesia, ID

r

•

H. A. W1sesa, Universitas Indonesia, ID

•

H. R. Sanabila, Universitas Indonesia, ID

•

K. M. Kurniawan, Universitas Indonesia, ID

•

M. Mega, Universitas Indonesia, ID

•

M. Ni'ma, Universitas Indonesia, ID

IC'.\< 'SIS '21111

I

t

r

ICACSIS 2014

'

-.

f

•

•

M. M. Soleh , Univers1tas Indonesia, ID I. Suryani, Universitas Indonesia, ID

1 _{M. Prav1tasar1, Univers1tas Indonesia, ID}

•

I

•

N Rahmah, Un1vers1tas Indonesia , ID

•

Putut, Univers1tas Indonesia, ID

t

•

R. Musf1kar, Untvers1tas Indonesia , ID

l

•

R. Latifah, Universitas Indonesia, ID R. P. Rangkuti, Univers1tas Indonesia, ID

•

V. Ayumi, Unlvers1tas I ndonesia, ID

J

ｾ＠

_ｾ＠

5

ｾ＠

l

I

J

l

\

I

ｾ＠

L

ｾ＠

f

l

.

ｾ＠

i

l

•

i

ｾ＠

'

I

ｾ＠

l

f

1

I

'

l·

l

1

ｾ＠

i

ｾ＠

t

t

l

-- -

-- - --

- - ----

·

I('.\ ( 'S IS :!II I I

--- --

-CONFERENCE INFORMATION

Dates

Organizer

Venue

Official language

Secretariat

October 18:h (Saturday) - October ＱＹＧ ｾ＠ (Sunday) 2014

Faculty of Computer Science, Un1versitas Indonesia

Department of Computer Science, lnst1tut Pertanian Bogor

Faculty of Information Technology, Un1versitas Tarumanegara

Ambhara Hotel

Jalan lskandarsyah Raya No. 1, Jakarta Selatan, OKI Jakarta, 12160, Indonesia

Phone : +62-21-2700 888

Fax : +62-21 -2700 215

Website : http://www.ambharahotel.com/

English

Faculty of Computer Science, Un1versitas Indonesia

Kampus UI Depok

Depok, 16424

Indonesia

T: +62 21786 3419 ext. 3225

F: +62 21 786 3415

E: icacsis@cs.ui.ac.id

W: http://www.cs.ui .ac.id

Conference Website http://icacsis.cs.ui.ac.id

v

( '

IC:\C

Ｇｓ ｉ ｾ＠

2111 .1

PROGRAM SCHEDULE

· Saturday, October 1s1t1, 2014-CONFERENCE

Time ' Event

·-

- -

.

-08.00·09.00

I

I

09 00-09.30

!

Opening

I

I

I

I

I

09.30-10.15

I

Plenary Speech I

I

I

' 10.15-10.30

1 10.30-11.15 Plenary Speech II

1115.12.30

12.30-14 .00 Parallel Session I :

Four Parallel Sessions

1 14.00·15 30 Parallel Session II:

Four Parallel Sessions

15.30-16.00

1 16.00-17.30 Pa rallel Session Ill :

Four Parallel Sessions

ｾ＠ 17.30-19.00

1 19.00-22.00 Gala Dinner

i

-

- -

-Event Details Registrat ion

I

Rooms

- -

ｾ ＭＭ ＭＭＭｾＭＭＭＭＭ ＭＭＮＮＮＮＮＮｬ＠

Ope ning rom t h D

e

ean o Faculty of

j

Computer Science Universitas I

Indonesia/General Chair of ICACSIS I 2014

1 _{Dr . Ir. Basuki Yusuf lskandar, MA} Dirgantara

I from Ministry of Communication and Room, 2"° Floor

I

Information

I

Coffee Break

ｾｲｯｦ

Ｎ＠

Dame Wendy Hall

I

from Southampton University, UK

Lunch

See Technical (Parallel Session I Elang, Kasuari,

Schedule) Merak, Cend rawasih

Room, Lobby Level

See Technical (Parallel Session II Elang, Kasuari,

Schedule) Merak, Cend rawasih

Room, Lobby Level

Coffee Break

See Technical (Parallel Session Ill Elang, Kasuari,

Schedule) Merak, Cendrawasih

Room, Lobby Level

Break _I

Dinner, accompanied by music Dirgantara Room,

I

performance and traditional dances 2nd Floor

I

l

l

I

:

J

I (':\(

'SIS

:?11 1.1

• • Ｑｾ＠ JCAC515

2014

-

- ·

-I

Sunday, October 19th, 2014-CONFERENCE

I

Time

r

08.00-09 00

Event

09.00-10.00 Plenary Speech Ill

I

10.00-10.15

L

10.1s-11.3o Plenary Speech IV

11.30-12.30

I

Parallel Session IV :

1230-14.00

Four Parallel Sessions

Parallel Session V : 14.00-15.30

Four Parallel Sessions

15.30-16.00

1 16.00-16.30 Closing Ceremony

Event Details

Registration

Ors. Harry Waluyo, M .Hum

1 from Directorate General of Media,

Design. Science & Technology Based

I

Creative Economy

Coffee Break

Prof. Masatoshi Ishikawa

from Un1vers1ty of Tokyo, JP

Lunch

See Technical (Parallel Session IV Schedule)

See Technical (Parallel Session V Schedule)

Coffee Break

J Awards Announcement and Photo

Session

-- -

-Rooms

Dirgantara Room, 2"d Floor

Elang, Kasuari, Merak, Cendrawasih Room, Lobby Level

Elang, Kasuari, Merak, Cendrawasih Room, Lobby Level

Oirgantara Room,

ind Floor

XI

ISBI'\ : 978-979-1-121-225

'

l

t

.

•

[('\('SIS

ｾｬｬｊ＠

I

Table of Contents

\\ drnmc

ｾ＠ｫｳｾ｡ｧ｣Ｚ＠

from

Gcnc:ral

ｃｨ｡ｩｲｾ＠

ｗ･ｫｯｭｾ＠

ivkssagc

from Dean of l·::ic.:ull) nf Cumpuler S1.:ic111..:e Uni\.ersi l) of Indonesia

Committees

Confrrc:ncc Information

Program S<.:hetlule

l'ahk olTontcnb

Comptltl'r Netwo r ks, A rc hitecture

&

lligh Perform a nce Computing

ｾｬｵｬｴｩｷｲ｣＠

Computation of Tacti<.:al lntcgrntiun System in the Maritime Patrol l\ircrall

ｵｾｩｮｧ＠

lntd Threading Building Block

.\/11hu111111ud

Furis Fathoni.

Bumbang

Sridadi

ｬｭｲｮｴＡｲｳｩ｜ｾ＠

Virtual 30 Environment based on 499 fps Hand Gc-,1urc Interface

.\/11/wmmacl

Sakti Afrissalim

Im pro\ e fault tolerance in cell -based evolve hardware nrchitccturt!

Chemin

IVnngyai

A New Pntients· Rights Oriented

Modd of EMR Access Sccurit)

}'B D1ri

Setiantn. >'ustina Retno

IV.

Utm11i

Element Extraction and Evaluation of Packaging Design using Computational Kansei

Engineering Approac.:h

Tat!fik

Djutna. Fajar .\f1111ichp11tra1110 . .\'i11u llairiyalt. t:(/ira Febriani

lmcgratcd Information S)stern Specification to Support SSTC

Allmacl Tamimi

Fadhilah. fudlro

Giri

Sucahyo. Denny

A Real Time Simulation Model Of Production System Of GI) ccrol Ester With Self

111

v

IX

x

xiii

7

13

19

33

Optilnization

39

hrnn

Aang Soenandi, Taujik Djallw

-

- -

--

-

Ｍ ｾ＠

-

- -

·

[( '..\C

'SIS

'21ll I

I

)c\ ､ｯｰＱＱＱｾｮＱ＠ or'

l ni\ l.'r-.il)

or

ｬｮｊｯｮ｣ｾｩ｡＠ ＺＭＭＮｴＺｾｬ＠

( i1.·n1:rntil)ll I in.'" .ill PrnWt>

pc

.tnd

ＮＮ｜ｮＺ｣ｾｾ＠

Control \\ ith

Deep

P:.1ckc1

ｬｮｾｰｬＮＧＱＮＧｬｩｯｮ＠

llan"1

.\lllh1111111wd \t1::1l'/. /'01111_\ .·ld/11 St1h1/\f i1111 .·ll/1111 /'rnekal. (j/1111/11 (111tmldi11

Reliabk

Data

ｬｫｬｩ｜･ｲｾ＠

:'vkd1anism on Irrigation Moni toring S}'>ll.'m

.f111wicl1

H11Ji Sa11gl!r. fler11

Sukoco.

Satyr111fo Su1>1011w

E-<;ovcrnm ent

1:-

valuation on

People A

spe<.:t in

K

nn" lctlgt: Management

S

ｾｳｴ･ｭ＠

I mplcmcntation :

J\

Casi.'

Stud)

or

Bank

I

mloncsia

llandrf!

Durianu. lclu

.·I\ 11

Tri.\·110111_1.

and

/>11111

ll'uri llamlaycmi

( im

ernment

Kno\\

ledge Managcmenl

ｓｾ＠

s1em

Analysis :

Case StuJy

Badan

Kepl!g<l\\aian Negara

Eli11

(

'ahyaningsih. So/iyunti !11dria.wri. Pinkie ,·lnggia.

/Jww /ndra

Se11s11se.

Wa'1y11 Ca1111· Wihowo

ｓｨｾｭＺ､＠

Scr\. icc in E-GO\ernment Sector: Case

SIUU)

or

Implementation

in

Developed

ｃｯｵｮＱｲｩ｣ｾ＠

Raviku I lc{/i=i.

Suruya

,\ fiskon. A::i::ah Ahtlul Rahman

GIS-based DSS in e-Livestock Indonesia

Arie

I

Rcmwcllum.

Dona

/nclra

Se11.,·11se . • I

/uhammad

Ocwviww

f>ralama.

l'ina

A,rumi. :lniati ,\/urni

ａｲｹｭｵｲｴｬｾｲ＠

Influence

Or

Presidential Candidates E-Campaign To\\ an.ls Voter" In

201-l

Presidential

Election In Repuhlic Of Indonesia

ra11i

i\'11rlwcl1J·a11i.

Al/Cmg Kurnia.

lnyml

Satria

Information Sccurit)

Risk

Management Planning:

J\

Case Stud)

at Application Module

51

55

65

73

81

87

of

State Asset Directorate General of State

Assc1

Ministry of Finance

93

Sigit

Praseryo.

Yudho Giri Sucahyo

I'

Campaign

2.0:

Analysis of Social Media Utilintion in 201-l Jakarta Legislative Election

99

Deem

.

lpricma

Ramadlw11

ｾｉ＠

55

SS

73

31

)3

)9

IC'.-\('SIS

ｾｬｬｬ＠

I

TO\\artfs

ｍ｡ｴｵｲｩＱｾ＠

:-.todd

for E-(iO\l:rnmcnt lmpkmt:ntation Basl!J on Suc\.'.ess !·actors

105

i>11r111c11rc111

/Jt1I!.111da

.\"api

t

11p11!t1

The

Cri1i1.:al

Succl!ss Fal·IOrs

10 ｄ｣｜ｾｬｵｰ＠

an lntq?.rall.:d

Applic11iun of runa f-i,hing Data

\lanagl'llh!lll

in

Indonesia

l>e\'i Fi1riwwh. .\'ursiclik Ht!rtt J'rapto11" . . frlmwcl .\'i:"r I liday"11w .. lniali ,\lumi

.·/ 1)'111 Ill'/

hy

A Com:ep1ual Paper on ICT as National Strat1:gic Resourcl!s t<mard National

Compl!titivcness

l {

13asuki

Yusuf

lskandar i.lnd h:idhilah

Mathar

Ba.rnki

r11s1!l lskcmclar

e111cl

Fm/hi/uh .\!a1lwr

Enterprise Computing

Qualil) l:-.valu:ition

of

Airline's E-Commcrcc Website!. A Case! Study

or

AirAsia and

Lion Air Wehsitcs

Famh Slwj/ra q/1mdi. Jku A

l/i11a

Hotspot Clustering l hi ng DBSCA N Algorithm anti Shin} Web Frame\\Ork

Karli11a

Khiyarin Nisu

Analysis

of

Trust Presence Within

E-Commcrce

Websites:

A

Study

of

Indonesian

E-Commerce Websites

M11ha111111ucl

Rijki Shilwb.

Sri

Jl 'ahy1111i .. -ll111wd .Vi;ar Hiclaya1110

The

Impact of Customer Kno\\ ledge Acquisition Lo

KnO\\

ledge

Management

Benefits:

A Case Study in lndonc.:sian Banking anc.J lnsurnnce lnc.Justric.:s

.Huhammad R[fki Shihah. A.Jeng Anugrah Leslari

111

117

123

127

131

137

A

System Analysis and Design for Sorghum Based Nano- Composite Film Production

143

Belladini Lovely.

Ta1!fik pja111a

Moving Towards

PCI

DSS

3.0

Compliance:

A

Case Study

of

Credit Card Data Security

Audit in an

Online

Payment Company

Muhammad Rijki Shihab, Fehria11a .\/isdia111i

- -- - - - -- - -

- -- - -

- - -·

-149

I( ＧＮ ｜＼ＮＧｓｉｾ＠ .!lll-l

:\n ﾷ｜ｮ｡ｬ＾Ａ｜ｩｾ＠ and Dc-.ign ol I ｲﾫｈＺ｣｡｢ｩｬｩｴｾ＠ In I ro1cnVannamc '-;hrimp PrnJuct ｨ｡ｾ｣､＠ l"1

1)1g11al ｬＳｵＡＭ＾ｩｮ｣Ｂｾ＠ ｬＺＺＮ｣ｯＺＺＮｾｾｴ｣ｭ＠

li111/ik

IJjo111a 11ml ..ldi1w (ii111111111k11

ｂ｡ＩｾＺＮｩ｡ｮ＠ Rough Set i\todcl in I ｬ ｾ｢ｲｩ､＠ Kan sl.!i Engineering for Ht:'vcragc Packaging

Oc ... ign

..l:r(/inl'(lll um/

Tau{tk Dja11w

Predicting Sma11 • lome Lighting Ocha\ iour from Sensor::. und User Input using Ver)

155

163

Fast Dt!cision Tree with K ernel Dt:n sity Estimation and Improved Laplace Correction

169

Ida Bagu., f>wu

Pl!rmlnyu Di11ottt. and Boh f/orclia11

Vi-;ual Llsahilit) Design for Mohilc Application Ba!->cd On lJscr Personality

175

Rirn. lk111·ia.

Tc11!fik Dja11w.

am/ foni

.\'11rhudrya11i

Forma l Method Software Engineering

Interacti on bl.!mccn users and builJi ngs: ｲ｣ｾｵｬｴＡ｜＠ of a multicrdcria anal) Sis

181

Audrey Bona

u11d .Jew1-.Wurc

Salo/Ii

Digital \'vatermark ing in audio for copyright protection

187

Hcmis

M11s1apho

ancl Houdraa Bachir

An Extension o f Petri Nct\\Ork for Multi-Agent System Ri:rresenli.ltion

193

Pierre Sam·oge

Enhancing Re l iability of Feature Modeling" ith Transforming Representation into

Abstract Behavioral Specification (A B S) 199

.\lulwmmod lr/(111

Fadhilluh

:vtaking "Energy-saving Strategies": Using a Cue Offering Interface

205

, >'as111aka

Kishi.

Kyoko

Ito. and Shogo .\'isltida

Extending V-modcl practices to support SRE to build Secure Web Application

111

.-1/a Ali ..tbdulm=eg

'

'-'

II

>3

)9

15

11

Ｉ ﾷＩｾ＠

.

__

.)

r

I

f

H ':\('SIS :?Ul -1

!

I

t

t

l

l

I

f

•

i

•

t :

I

l

_l

f

l

ｾ＠

t

t

i

f

'

I

i

t

.

i

I

f

t

t

I

t

1. ;

I

Social Nl!l\\ or!- Anal)

ｾｩｳ＠

for P..:opk "ith Systemic

ｌｵｰｵｾ＠

tr) thcrnatosus

ｵｾｩｮｧ＠

R4

I·

ramc \\

ork

.·lri11f..."ar/11101111d F1rmc111 .-lrdiall\\<1/i

｛ｾｰＱＮＭｲｩＮＮＺｮ｣ＮＮＺｳ＠

l 1-;ing /2St\I.

,\/uriu U(/uh Sirl!gar. John Derrick. Siobhan North, and

An1/10ny

J.

H.

Simons

Information Retric\•al

Rt!latiH:

ｏ｣ｮｾｩｬＩ＠

Estimation using Self-Organizing

ｍ｡ｰｾ＠

Denny

Creating Bahasa Indonesian - Javanese Parallel Corpora Usi ng Wikipcdia Articles

Buyu Di.\t

imran

ｔｲｩ Ｎ｜Ｂ ﾣＧﾣｾ｜＠

·a

Classi fo:alion

of

Campus E-Complaint

ｏｾ｣ｵｲｮ･ｮｴｳ＠

using Directed Acyclic

Graph

Multi-217

223

23 1

237

Class

SVM

Based on Anal) tic Hierarch) Process

245

Imam C''10/issodi11 . .\loya l\11r11imrnti. lndriati, and ls.\·a A

n1

a11i

Frame\\ Ork Model of Sustainab le Supply Chain Risk for Uairy Agroindust ry Based on

Knowledge Base

Winnie Septiani. ,\farimin. Yeni llerJ(reni. mu/ Ue.\betini

ｈｵｊｩｾｪ｡ｲｯｫｯ＠

Ph)sicians· lnvolvem1.:nt in Social Mcuia on Dissemination of Health Information

Pmd I

hrahim Nainggolan

A

Spatial Decision Tree based on Topological Relationships for Classifying Hotspot

Occurenccs in Bengkalis Riau lnuoncsia

raumil Miss

ｬ｜ｨｯｩｲｾｲｮｨＬ＠

and Imus Sukaesih Sita11ggw1g

Shalla\\' Parsing Natural Language Processing Implementation for Intelligent Automatic

253

259

265

Customer Ser" i<.:e System

271

.

lhmad f:ries

.-I

mares ..

·ldhi l\m11adi. and .\'i .\fade SutYika J.nrari

SMOTE-Out, SMOTE-Cosine. and Selected-SMOTE: An Enhancement Strategy to

Handle Imbalance in Data Level

-

-277

I<

ＧＺ ｜ｃＧｾｉｾ＠ ﾷｾＱＱＱＱ＠

Fajri

J\oro

r\Japti\c

lnlorrn:.nion

L-::-.tra1:tilHl

of Di-.a-.1cr

Informat ion

from

1 \\ itt\.·r

Ralph l"incent .I Regulodo . .Je11i110

L.

Chua . .Jmrin

L.

Co. I

h:r111w1

< ·

Cht:ng

Jr ..

. ·lngl!!o l/rul'I.' I . .

\fagpanruy. am/

J\n.\fme .\la Dommicfllt' F

A.'a/011

Citatit111

ｾ｣ｮｴｩＺｮ｣｣＠

ldcntilic;.nion anc.I Classification for Rdatc:d

\\.'ork

Summari1ation

D1ri

llt!mlrutmu

Jl'idnmtoro. am/ /111od11dtli11 .·l111i11

Experiments

on K c)\\

ord

ｬＮｩｾｴ＠

Gem:ration

fl)

T

crm Distribution Clustering for Social

ｾＸＳ＠

287

\kdia Dma Classilication

293

ll"tl\(}I/

roncla

Clllll ' '·' '"

/'11nrnria11ti

Tourism Rec.:ummcndation ｂｮｾ｣､＠ on Vector

Space Model and

Social Rc:comml!ndcr

Using

Composite Socia l Media Extrm:tion

299

H11.rn11/

J\lwtimah.

1

ul!fik

f)jat11a.

and Yam

.\'urluub:nmi

Learning to Rank for Dctc:rmining Relevant Document in Indonesian-English Cross

L:mguagc: Information Retrieval using IJM25

.\\cmclra

Sari and ;\Jirna

A

driani

Pattern Recognition

&

Image Processing

Forccasting thc Length of the Rainy Season Using fimt: Delay Neural Net\\ork

Ag11.'> Buono .. \111/wmmad

:b:l'lwr

:lgmalam. and Amllliu Fitrcmty

Almira

11) brid Sampling for Multiclass Imbalanced Problem: Case Study of Students'

Performance Pred iction

lrantlwnee l'rachuabsupakij and

.\'11r111wa11

Soontlwmphisc!i

:-..Julti-Cirid ·1 ransformation for Medical Im age Registration

Por<l\rat

Visutsak

\Jock I Prediction for Accred itation of Public

Junior I

ligh School in Bogor Using Spatial

Decision Tree

ｆＮＱＱ､ｷＱｾ＠

.

Pumama

Ciiri and Aniati .\/umi

.·Jnwurth\' _{. .}

305

311

3 17

..,...,.,

->-.:>

329

X\·jjj

ISBN : 978-979-1-421-225

t

,

.

.,.

t

I.

I.

I

1

t

!

Applil.'.ation of

ｬｫＱＮＺｩ ｾｩｯｮ＠

rn.!t:

ｃｬ｡ｳｾｩｬｩｴＺｲ＠

for Singh.: :'\udcotidc

ｐｯｬＩｭｯｲｰｨｩｾｭ＠

DisCO\l!r)

from Nt::\t-Gc.:m:ration St:quc.:ncing Data

.\f11ht1111111t1d

.·lhmr

/,·1ioc/1. ll'irn11

.·lmmta

f.:11s11m<1.

untf I .\fadl' I

mmo

Altemuth c

I

c.:.iturc.: btraction from D1gitizc.:J

ｉｭ｡ｧ･ＮＺｾ＠

of

ｄｾ｣ＺＺ＠

Solution')

tb

a

ｾｬｯ､｣ＮＺｉ＠

for

335

Algal Bloom Remote Sensing

3-l I

Ro}!.er

l.11i.\

C)-.

.Jvi.:l Jlao.

Frie P1111:<1/a11.

wul !'ram!

.\farit'I

011g

Iris Localization using Gradient

Magnitude nntl l·ourier Oc.:s1.:nptor

Stell art

.\'e111ww<! .• -11110

S .\'ugrolw. Reggw .\'I lurtmw ..

\/ol1e1111macl l '/iniam:rah.

um/

Ｌ ｜Ｏ ･ｩｴｾｲ＠

l.ayoouri

Multiscale Fractal Dimension Modelling on Leaf Venation Topo log) Pattern of

Indonesian Medicinnl Plants

A:i;

Rahmad

re11i

Ｏｬ･ｲ｣ｬｾｲ｣ｮｩＮ＠

Agus Buono.

and Stephane

ｄｯｵ｡｣ｾ｜ Ｇ＠

347

353

Fuzzy C-Mcans for Deforestratiun Iden ti

Ii

cation Basetl on Rc::mote Sensing Image

359

Setia Darmm1•<111 .·Uandi. J'eni

ｈ･ｲ｣ｬｾｲ｣ＱＱｩＮ＠

and l.ilik 8 Prasetyo

Quantitative Evaluation for Simple Sc.:gmentation SVM in Landscape lmagt.:

£11cla11x

f' unwmu (j iri

and A

11ia1

i ,\-111r11i

A

ry11111r1

'1y

ldt.:milication of Single Nucleotide Pol)inorphism using Support Vector Mm:hinc on

Imbalanced Data

Lai/an Sa'1ri11a flasibuan

Development of Interaction and Orientation Method in Welding Simulator for Welding

Training Using /\ugmenetcd Reality

.·lrio

Swwr Baskoro . .\lohammml rl;:1rar Amat. and

ｒ｡ｮ｣ｾｲ＠

Pa11ges111 Ku.nrunv

Tracking Efficiency Measurement of Dynamic Models on Geometric Particle Filter

365

371

377

using Kl.D-Rcsampling

38 I

.-l/e.w11der

..I S Ci1111mrn11. ll"is1111

.Ja1111iko. l'ektor

De11·c11110.

F.

Rad111wdi. und

F

Jo\'lm ,

i\onlinearl) Wc.:ighted Multiple Kernel Learning for Time Series Forecasting

Agus Widodo. Indra Budi. and Bellllrati Widjaja

385

:

I ( ' :\ ( 'SIS :20 I I

·

-D1:.!0rllllll ｜ｮ｡ｬＩｾｩＮＬ＠

nr

I liaarchical \li:-.ing I c:chniquc:

Oil ｾｉｐｉ＠ {1 <...tt1T\l l11ld

'itanJ;m.1

lkl11rw1t1

1:../fitri .. \11111111'1 .\/11/wrum. am/ ,\lulwmmcul S/10h1rt11

.\ C\1mpari son of 8ackpropagatinn anJ 1.VQ: a t:a'>c:

ＭＬ Ｑｵｊｾ＠

nl. lung

ｾｮｵｮｊ＠

recognition

Foe/hi/ah Sra/ria .

.

.

·I i:11\

811onv. a11d

Hih l'aru/111111 Silalahi

. '

:\rcPSO:

ｌｬｬｩｰｾ･＠

Di.:tcctiun

ｾｫｴｨｯ､＠ ｵｾｩｮｧ＠

P:.irtidc

ｾ｜｜｡ｲｭ＠

Optimization and Arc

Combinat ion

..lpri11e1/di. lkh\Clllll! Huhihl(•. Roht'th Rul111wt11/lah.

a11cl

..r

K11miC111 u11

31)

Virtual Pet

Game

ﾷﾷｾ Ａ ｯ｡ｲﾷﾷ＠

With .t\l1gmcnti.:J Rcalit'

v • 10

Simulnte Pct

ｒ｡ｩｳｩｮｾ＠ v

S1.:cnario

on

Mobi

k

De'

i1:c

(

'/Ufen Allen,

Ｎｊ･｡ＱＱＱｾＱ ﾷ＠ J1

rugc111tha.

ancl DC1ri11., ..lnclww Haris

J\utornatic

Fcial

Organs Segmentation Using Multila) er Super Pixel and Image Moment

Feature

R. Rohma1111/uh .. \/. Anwar

J\l(l.\"/1/11,

Apri11C1ltli.P. i\111r.•w1110.

n.

ll'iweko.

C//lU Herl)·

Integration

of

Smoke

ｄｩ ｳ ｰｾｲｳｩｯｮ＠

MoJeling with Earth·

s

Surface Images

..I

Sulaimon. M Scully

Performance

of

Robust Two-dimensional Principal Component for Cl3ssi fication

Diuh

£.

I lerwindiati, Semi,\/. Isa. und Ju11sv11

He11d1:i·li

Pareto Frontier Optimization in Soccer Simulation Using Normalizc:<l Normal

Constraint

Darius

.·lndana Haris

Fully Unsuperv ised Clustering in Nonlinear!) Separable Data Using Intelligent

Kernel

K-:\1eans

Te11y

I

fanda.vuni and Ito

Wasifo

Robust Discriminant Analysis for Classi lication

of Remote Sensing Data

r

Wina.

Dyah

£.

Henvindiati.

and

Semi.\/.

Isa

Automacic Fetal Organs Detection and Approximation In Ultrasound Image Using

391

403

409

415

423

429

437

445

Boosting Classifier and I louuh Transform

455

- v

f'

l

I

l

9

r

7

•

1

,.

i.

I

5

l

t

9

!

!

I

5

I(':\< Ｇｓｉｾ＠ ﾷｾｵ＠ 1.1

\I .·l1111w·

.\/0

1

.\11111. Wi,1111./01111iko

\I lc1ht1f

7illl

t1k11/ u11d !-uri' .

II.

I/ii

ｐ｡ｮｩ｣ｬｾ＠

\\\arm Optimation

ｨ｡ｳｾ､＠ ｾＭｄｩＱＱＱＱＮＮＺｮｳｩｴＱｮＬ､＠ ｒＮＱｮｊｴＩｭＱＱｾ､＠

I l\H1gh

Tran:--torm tor

l'ctal

I

ｬｾ｡､＠

13iomctr) Detection anJ r\ppro\llllJtion in l ltra'iOUllJ lm<1g ing

f'llfu Satll'iko. fkJ11w111l f/ahihie .. \/ .-11111w· .\la'rnm,. l11dn•cH

Fehrion. £11rico

B mlitmw

Oigital Library

&

Dista nce Learning

Gamified E-Lcarning Model

Bnse<..l

on Communit)

uf

lnqu ir)

:foci

i

k

a

)'i

u

11

w f."t

u 111 o.

.

.f./

U

(

1

..I

111ri'1111.

A Iha

111

Fi

kn A

11.

Fa

fl

11

Ro

'1111

uh

.\'I/

r

Wohiclah.

mu/

Kmiyah .\/.

.l111111s

KnO\\ ledge Management System Development '' ich Evaluation Method in Lesson Study

463

469

Activity

477

.\lure in .\Ii

ks

Cl .\ lort!hit1 ..

·1

r111l'i11

ZR.

l.ang

i. wul

foanes BC1ncl1111g

Designing Minahasa Tou lour 30 Animation Movie

as

Part of Indonesian e-Cultural

Heritage and Natural JI istory

483

Sum/ey

Karotnr

Learning Content Pcrsonali?.ation Based on Triple-Factor Learning Type Approach in

E-learning

489

,\lira Suryani.

Hany

Budi Santoso. anti 7ainal A . llasih11a11

Adaptation

of

Composite E-1.earning

ｃｯｮｴ･ｮｴｾ＠

for Reus3blc in Sm:irtphone Based

Learning System

497

liermw1 Tolle.

l\ohei Arui.

and

.-I

1yo

Pi11c11ulito

The Case Study or Using GIS as Instrument for Preserving Javanese Culture in a

Traditional Coastal Batik. Indonesia

503

l}i beng .lap am/ Sri Tiatri

IC'. ｜ｃｓｉｾ＠ 2111 I

A Comparison of Backpropagation and LVQ · a case

study of lung sound recognition

I .1dhdah ｾｾ［ｩｦｲＱ｡ Ｑ＠• .-\g.11' 1\111111<1 and Bib ｐｾＱＱｵｨＱＱＱＱＱ＠ \1lalahi ;

!>"f't1r1111e1/f o)

c

011/11/llttl'r

s,

'''"' ,.

t111d ｬｫＯ ＾＼ＱｲｬｬＡｭｾﾷＱＱ ＱＱＱＯ＠ _\lt1tht·ma11c:!>

Fm·u//\ of _\/01J11:11111t1C\ and .\'a111rcil ｓ｣ｩ･ＱＱｵｾ｜＠ Ｏｊｵｾｯｲ Ｎ＠ lj.!rtcttlturul {_ Ｇｮｩｮﾷｲｾｩｴｹ＠ Bog or. /ndune :!>iu

l·.mail: 1

fadhiinlh) afria2

_{I 02}1_{1/ gmail.1.:0111.}

'pudc:;ha

_<t·) ahoo.<.:o. id . 1bibparuhum I

'it.yahoo.com

.11>,11-.1<1 One \l:l) to C\aluatc 1h.: '>lak 1lf 1he

lungs 1s b) ｬｩｾｴ･ｮｩｮｧ＠ ll' breath sounds U'>ing

stethoscope I his techni4uc is !..no\\ n us auscuhat1on.

This technique is foirl) simple and inc:-.penmc. but ii

has some disad\antagc. 1 he) arc the rec;ults

or

subjecti\c anal) !>is. human hearing is le.,, -.cn,ili\e 10 lo\1 fre4ucn<.:). en' ironrnental noise and paucrn 11f

lung ｾｯｵｮ､ｳ＠ th:it alnllhl '>i111il:1r 13ccau'c tif' thc,e

ｦ｡｣ｬｬＱｲｾＮ＠ 1111sd iagno:.is cnn occur if proccdurc of

auscultation is not done properly. In this research.'' 111

be made a model of lung sound recognition \\ ith neural network approach. Artificial 111.:ural networl.

method used ii. D:icl.propagation (UP) and learning

Veclllr Quan11.zatit)n (L VQ). Comparison

or

these t\\O

methods ｲｾﾷｲｦｯｲｭ｣､＠ ICl determine and recommend

algorithms which provide better recognition accuraq of speech recognition in \he case of lung sounds. In

addition 10 the above two methods, the method of Mel

Frequency Cepstrum Coefficient (MFCC) is also used as method of feature extraction. The results sho'' the

accurnc) of using Bacl•propagation is 9:>.17°0. \1hile

the value of u:.ing 1he L VQ is 86.88° o. It can be

concluded that the introduction of lung sounds usmg Backpropagation method gives beuer performance compared to the LVQ method for speech recognition

cases of lung sounds.

I. I 'J fl{!)()\ 'l 1101'

Lung ｳｯｵｮ､ｾ＠ are pan of the respirator) sounds.

Respiratol) sounds include the sound of the mouth and trachea while lung sounds occur around the chest. Sounds of lungs occurs due 10 air turbulence 11hen the air enters the respiratory tract during breathing. This turbulence occurs because the air flow from the air duct that is "ider 10 narrower airways or ,·ice versa.

In general. the sound of the lung is divided into rwo. sounds is the sound of the lungs that nre not detectable respiratory abnormality. whereas abnormal lung

;397

sounds ;ire s11lllldS lung dborder.

On.: ＧＧ｡ｾ＠ hl e' alua\c the state of the lungs 1s b)

ｦｯＮｴ･ｮｩｮｾ＠ 10 hrcath .,oumh using a stethoscope 1 his

tc::chnique is ｾｮｯＢｮ＠ as auscultation ｴ･｣ｨｮｩｱｵ･ｾＮ＠

Auscultation technique 1s a basic technique that 1'> u-,ed

b) ph)'sic1ans to C\ aluate breath sounds ｬＢｨｩｾ＠

tedmique ｩｾ＠ fairly -,impk and rnexpensive. hut it has

the tfr,ath antnge that 1he results of a ＭＮｵｾｪ･｣ｴｩ｜｣＠

Ｚｩｮ｡ｬｾＧｩｩＧ＾＠ ＬｾＬ Ｎ＠ I he ｮＺｾｵｨｳ＠ Llf the ｡ｲｲｮｬｾ ｳｩｳ＠

ur

hrcalh

:.om1L1' ｵｾｩｮｧ＠ <lll"cuh<11ion technique relics 1ln the

ability. auditor)·. and experience of the docll1r \\hO performed thc :urn l) sis. In addition. human henring is

less scnsllive 111 lo" frequcnc) sound is ｡ｬｾｯ＠ a problem

in this technique. hccnusc the sound of hrcathing

occupies a lo\\ enough frequency. The nc:-.1 prnhkm in auscullatiL'll technique<; are the prohl.:111 111° en\ imnmenrnl noise nnd the pallern of sou nds that almost similar between one 1ype of breath sounds and the other. Because of these factors, misdiagnosis can occur if the procedure of auscultation is not done properly.

Based on the abO\e problems.

''e

interested in

conducting rc!icarch in performing speech recognition of normal lung and lung sounds \\ere detci.:h:d

interference (abnormnl). Sounds of lungs resuhing in

some ca!.es of the disease showed a specific pattern

1ha1 can he n:cogni/ed. This soun<l paltern can b.:

1aken as materi:il for diagnosis

131.

Lung sound will be

classified into four ､｡ｾｳＮＺｳＮ＠ namely tracheal. "esiculnr.

cradde am.I \\hCe/e Tracheal and vesicular ｾｯｵｮ､＠

ｩｮｊｩ｣｡ｨＡｾ＠ nomal lung. "h1lc the crack le and \\ hcete

sound indicates abnormalities in the lungs

l4Jl5 ].

Voice recognition is an effort in order to the voice

can be recognized or identified so that it can be used. A \Oice recognition S)Stem is a computational application that is able to identify or verit) the ::.ound is autonrnticnlly. These systems need to be trained in advance 10 be able to recognize the voice and the voice classifies according to the class. For such purposes. an algorithm is needed to train the spced1

recognition S) stem 10 perform more precise.

/\ ｳｾｾＱ｣ｭ＠ can learn and recognize pauems ol sound

•

I< '.\t'SI' ｾｬｬｬ＠ 1

-:.111 he: d1•11c: \\llh ｴｨｾﾷ＠ .1rpl11.:at11111 111 .1r111il.1<il ni:ur.il

ＱＱＱＮＺＱＱＱＱｈＢＭｾ＠ t\rtitkial n..:ura l 111:111ork i) a ..:.1111r111a1w11al

m..:thod that m1mu:) the: biolog11.:al 11c:11ral lld\\urlo. 111

hum.111 that c:nrabl..: trainl·J to sol\..: prubl..:111, . In

g<'.nl'r<il. thc:rl.' .1rc: '><'.\ c:ral art1tici.1I nl.'11ral ＱＱ＼ｾＱＱＱＬ ｩ ｲｬＮＮ＠

al11Mrthll1' that can bt: .ippli..:J tu pallcrn rc:..:.>gnitiun

l

hi'

n:'>c::trd1

"'II

llS\.' l\111 art1li-:i.tl neural n1.·t\l••rl..

.il,:1•rnh111' ｴｨＮＺｾ＠ .ire: llt1l.l..rr11pag.1t111n 1111'1 :mJ

ｌ｣Ｚ｡ｲｮ ＱＱＱｾ＠ \ l'Ctur <)ua11111a1io11 (l.V<)1 B,11h 111c:tl11•J)

hJ\ ,. :1.h antngc' and di'><'\lh ant:lg..:' 1)1 c:.1..:h

C11mp.1ri-,11n .11" these l\lo mc:thtid-. ｰ｣ＺｲｬｾＱｲｭＮＮＺｊ＠ to

det..:rmine and recommend algorithm) "htch pro11dc: bcn..:r recognition accuracr uf specd1 ri:cugniti(1n in

the case: of lung sounds.

In addition to u-;ing an ariilkial 111.:ural nc1111.1r!,.

algt1ruhm in thi: classification proc..:,s. it 1<. Ill) le)'

impMtant in voice recognition is the feature e:1.tractiun.

In tllllt ｳｴｵ､ｾＮ＠ foaturc e.\traction methods arc ｵｾＮＮＺ､＠ Md

he4uenc:) Cepstrurn Coenisient ( M FCC). MFCC 1s a

teaturi: e:1.traction method "hich g1\'es excdlent ri:suhs in ｣ｬ｡Ｍ［ｳｩｬｾｩｮｧ＠ normal lung ｳｯｵｮ､ｾ＠ and whee.le ｓｌｬｵｮｊｾ＠

16).

II. ｒｴﾷｾｬ ﾷａ ｒｬ＠ 11 Mt lltllll

Rl•:-c.1rch m..:thod .:onducted is ill11).tra1cd 111 I· igurc

I . In thl· image looks that re)earch bi:gin-; "ith J

fi 1..:rm11re revie\\ of thi: theories ｬｬｬＧ｣｣ｾｳ｡ｲＩ＠ tor the

comrletion of this study. such as lung suund thcor) .

M fCC. Bad: propagation. L VQ anJ MA TLA U

programming.

A .Hatawl.v Rt•.warch

Lung sound data obtained from the repositor) of

respirator) sounds on the Internet, namely Littmann

Repository. The sound data consist of J2 lung sound

data. "hich is divided into 8 tracheal sounds. 8

vesicular sounds. 8 crackle sound and 8 \\hCeLe sound .

Tracheal sound is a sound that is audible :11 the bai.e

of the neck and larynx. Tracheal sound is v..:ry loud

and htl- pitch i) high (Fig. 2a). <;o this sound is \W)

dear i;ounding compared to another normal lung. Inspiration and e:1.pira1ory are relatively equal in

length [7j.

Vesicular sounds arc normal breach sound:\ 1ha1 arc

hearJ on 1hc side ches1 and ｣ｨ･ｾｴ＠ near the stomach.

The ｾｯｵｮ､ｾ＠ ware ｾｮｦｩ＠ "ith a 1111\ pitch (Fig. 2b ). lnspira1ion sound) much stronger than e:1.piratol)

ｾｵ ｵｮｊＭ［ Ｎ＠ Often the c\piratol) proCl!S!' is hardl) a1H.lihle 17].

Cradles sounds arc a short burst sound chat is discontinuous (Fig. 2c), it is generally more audible

sound m the process of inspiration (8]. The conditions

that is causl! the cra ckles are ARDS. as1hm11.

bro11d11ectus1J.' d1rv111c hro11ch1t1s. comnlulo1io11.

ｴＡｭﾷｾＱﾷ＠ cHF. 111ers1i1iol lung disease dan p11/111011m)·

edema ( 9)_

-

- ---- ---

- -

- -

-

-398

•

•

1 ... .nu1 r "''tt•tto"

Mf(I

( \l.fb4 \f\ 1 h.f' tnoctt"t

ｻＸＮ｡＼ ｬｰｦｾｾﾷ ＧﾰＢ＠ ;ar.d l'VQ,

•

. .... , :w•J ..

1rf.1".-•.itC

ＬＬＬｾ＠ ﾷ ｾﾷ＼ＬＮＺ Ｌ＠

,,..,.'(; •"" ..

..-a '·"· Ｇｖｾ｢＠

y

f,,1,.,ct,;..i·\•tri .,. .. , S.:.'-.. J

r

y

Ooc:Vf'M ｮｬｾｨｏｴＧ｜＠ ｾｮ､＠

lfCIQtU

ll'hee:i>s sound is a type of sound that is continuous, have a hiuh pitch (Fig. 2d). and more often heard ori

the

･ｸｰｩｲ｡ｾｯｲｹＮ＠

This sound occurs when the Oo\\ of air:

throuuh the narrowed airnays due to secretions,'

forc::ign hod) or wounds that preclude (3J T ·

duration

or

the process of inspiration longt:r than lh

ＮＮＺＺＱＮｰｩｲ｡ｴｯｲｾ＠ proc.:ss. Besides that inspiring sound;

inh:n,,ity g rearer rhan the expiratory. The conditions .•

that is cause the 11'/iee:e are osth111a. CllF. chronic

bmnchiti.\. COPD dan p11/111n11ary edemu

f9J.

a. Tracheal b. Vesicular

d. Crackle d. Wheeze

Figure 2 ｔｾＱ＾｣＠ of tung sound (91

[image:21.632.336.556.79.489.2]

B .\t!glllL' l//cl/11)11

Lung )ound Jata 1h;i1 has been collected then 111ad1:

the c;ound -,1gnal :.egmentation ｰｲｯ｣｣ｾｳ Ｎ＠ This prnl.'.CS) i)

ｊｯｵｾﾷ＠ to cut '>l'Ulld 1nll> '>C\ era I )lllilllcr frame' 111 ••nkr

111 ･｡ＬＱＱｾ＠ pro..:cssetl Lung ) 01111J ｾｩｧｮ｡ｬ＠ '' tll 01: .:ut

baseJ on unc n:.,p1nllM) ＮＮＺ ｾ､･＠ SounJ sign.ti

'>egnwnt.1111111 pn•..:c)'> I\ d,,n,· ｨｾ＠ U!.lllg the A11Jaut>

wli"arc. llh1strm1011 o f ... ignal :.cgmentat1011 proc.:,.,

shtmn Ill Figure 3. ｾ｣ｧｭ･ｮｴ｡ｴｩｯｮ＠ pmcc ... s produce) 2-1

tracheal ... uund\. 2-1, c::.ic11lar sound-.. 2-1 ..:r.1cl.k :.t>unJ,

anJ ｾＮＮｪ＠ \\ hc1•ze -..oumh. ｾＱＱ＠ the total .,,1unJ, c>ht.11111:J ln1111 the 'e!-!1111·nt.it1o n prn..:c" I) CJ(1 "'und' data

C L>ata Di1trth111i1111 oj Trui11i11x am/ ｔｴｶｴｭｾ＠ Dura

In gc111!ral. then! are \\\O processes that occur in the

artificial neural network algorithm. namel)' training process and testing pn>ces). Sound data u!leJ in the learning process is called training data. while the sound data used in the testing process is calll!d h:::.t

darn. rhereforc. the lung sound dilta that lrn' heen

segmented di\'1ded into l\\O parts. nnmely the training

data and 1.::.1 data. The number of lung sounJ Jata i'

96 data. Data lung sounds Jrc then dclined ao, 80

training sounds and 16 testing !>Ounds data.

D .\/FCC F<'t1/11rc Extraction

MFC'C feature e>.traction is a technique used to

generate a \'Cctor which 1s used as an identilier

f

IOJ .

The feature is the cepstral col!fficient. cepstrnl coefficients that are used still considering the

perception of the human auditory system. MFCC

technique can represent a better signal than the LPC.

l.PCC and the other in speech recognition ( 11 ]. This is

due to the 11orkings or 1hc MFCC is based on the

frequenc)' difference that can be captured by the human car so that it can rcpresent how people reci:i\ c

sound signals [ 121. Figure 4 is a llow diagram of

MFCC.

=ot.itt '•'tct?r

•

ｦｲ ｾｲ＠ ..

ｾ＠

,,

...

(pp\ft Ut1'l ( \)...,f l( 1ru1 ' ... 'pn:,...,,

O''"''""' C'o\ .nf'

''""''°'"'

Figure -I Diag.ram ｯｦｾｉｆｃｃ＠

J , J\ t ' ｾ Ｉ＠.. ' ,,1 •

•' ﾷﾷｬｾｨＮ＾ ｴ ｬ＠t f t r)

"''

....

,

...

t h • ''('• l t.o •'l\\ i

\V'•l"l"'\i

I <'atur.: e\trnuiun l Ｑ ｾｲｮｧ＠ 1\1 FCC "111 not n:mm e

""> charactcrt)tics or 1mp0nant information of each

lung sound data. In add1t1on. the ;;11c of the lung sound

dat:i ｢･｣ｯｭ･ｾ＠ 11ut t1)0 large. ｕｲｯ｡＼ｬｬｾ＠ speaking. there

ar.: Ii\ c \tag6 o f ｾｦｦ ｃｃＮ＠ ic: frame block mg.

"indu" 111g. fast tuuricr tran ) fi.irm. md ｬｲ･Ｔｵ｣ｭＺＺｾ＠

'"app111g anJ ct>pstrum Cl>etlicient.

I ung ''·JUnJ data that ｨ｡ｾ＠ bl.'l'll ｾ･ｧｭ･ｮｴ･､＠ then

pc:rt,1rr11eJ tram.:- bf,1d .1111.!' Ｑｩｲｵ｣･ｾ ｩ＾Ｎ＠ Because the sound

dat.1 ha:. ｾ･ｧｭ･ｮｴＮＺｊ＠ u::.ing Judal:it) !>Olt\\are. frame

blo..:king ｰｲｯ｣｣ ｳｾ＠ "ill read the lilc') that have been

prc\ ｩｶｵｾｉｾ＠ -...:g111c1ttl.'d. Furthermore. "tndowing

ｰｮＱＮＺ＼ﾷｾＬ＠ using llamm111g W ind1H\ becau'c of c;imple

ll•rmula. I he 111.'\l ｾｨＺｰ＠ " the ra'>t 1-uuricr ·1 ransform l H

r ).

I· I

r

pruccs:. ｉｾ＠ used to COil\ at each frame that

h;i!) b.:.:n gcm:rated from thl! pre\ iou!) proce'>s from 1he

t 11nc dum:iin 111t1) th.: frcqucnc) domain, so it can be

more ce1si I) ob sen cJ Onci;: the signal is converted

from the time domain into the frequency domain, the

11e>.1 step is the mel frequency \Happing process, this

prm:ess need the tilter. thus will be formed M filler

before wrapping process 1s Jone. Nexl is the Discrelc

C:o!)inc rransform ( DC'Tl 10 get coeffisient cepsrrum.

thi::. Cocl1hicn1 cepstrum that 1s an output of the MFCC: [IJJ.

f ｓｩｭｩＯｷﾷｩｾｲ＠ .\ lt'cm11·e111c.:111 tCl11H ificu11011J

Lung sound data that has been obtained their

fraturc vector through feature extraction process will

he cla)osifo:d ｵｾｩｮｧ＠ t\\O method'> : bacpropagation and

L VQ. !"he rcsult:i of prelim1nar) proce'>sing of data in

the form ,)f voice ... ampk!) that ha'e been segmented

and th.: extracti1lf1 procc:is is carrieJ out \viii be the input ur lhe tv.o mcthutls.

Broadly speaking voic.: recognition lungs out b) using the back propagation and L VQ has t\\ o main pans, namely the stages of learning patterns (!raining)

and the stages of pattern recognition I similaril)

meCl'iurement (testing). Learning outcomes data

collected and stored as a learning model that can later be used to measure the similarity of a sound signal

entering the lungs ｾｯ＠ that the subsequent lung sound

signals can be recognized. Once the learning phase is complete, the next !>lage is the stage of pattern n:cognition (sunilant) measuremenrs).

I J Back.prvpog11111111

Backpropagation 1s a systematic method of artificial neural network using supervised learning algorithm and is l)·pically used by the perceptron with many la)ers (input la)er. hidden la)'er and output layer) to change the existing weights in the hidden layer [ 14}.

ＱＳ｡｣ｾｰｭｰ｡ｧ｡ｴｩｯｮ＠ is training which ｾｰ･＠ controlled

using "eights adjustment patterns to achieve the

minimum value of the error between the output of the

prediction results with rea l output. To get a network error. the forward propagation phase (fcedforward) must be done before.

..

. !

I

ll ＧＮ｜ＨＧＮＮＮＮ［ｉｾ＠ 11111

In th.: ｴｾ･ｊｴｵｲｮ＠ .1rJ pha<. .. • ea..:h un11 r\.'l:el\ ... , 111p111

!>1gnab and furnarJc; the signal 10 all 11111h in th .. · ｉｊｾ＠ er

,1bv' e 11 (the hiJJcn l.i: er). l:ach un11 111 the h1JJen

ｬ｡ｾ･ｲ＠ '\Umming the input signab that ii. ｲＮＺ｣｣ｬ｜｣ｾ Ｎ＠ 111 th1:-.

:.tcp a<:tl\at1un fum:tion is used to calculate the lllllput ... 1gnal

w

be ｾ｣ＱＱＱ＠ Ill .ill unit:. on 11 (lh1: uutput l.1: er).

I Ill> ... 1.:p d .. 111e

a'

111111.:h .h 1he m1111b.:r ,1f ｬｲｮｨｫＺｾ＠

la: er:- I hen. for each unit \.'I output -.ummmg the

vutput ［ＱｾＱＱＬｴｬ［＠ In th1) >tt."p .11.:tl\ at ion t'u1H..tt0n h al'>o I N'J I t' .:.i!..:ulatc the \.lUtput -.1i::nal

I

I.:;

I

Alier the foctlfornard phase. the next step 1!>

bad .. fornard pha)e In thi'> phase. ea..:h 1\utput unit

recei\e., a target pattern ｡ＺＮｓｬｬｏｾｬｬｴＮＢ､＠ ''1th the tr;1111111g

111put pattern and then .:akubt.: hi' .:rr11r informati<Hl

I hen c.1kulat .. • the corr.:1:t1l>ll ''eight'\ anJ baa.,

"'•rrn.th111 I f11, ... 1.:p June .h mu.:h ·" tho: n11111her .,f

lmld.:n

la:

er:-. l:a.:h hidJen unit Jelt.1 .. ummmg 111put:-.

(from unit> lrn;utcJ in th.: ｬ｡ｾ･ｲ＠ nbO\ e it). Multipl) th.:

Jella \alue b) the derivati\e of activation fu111.:tion Iv calculate error information. and then calculate th.: .:orrection \\Cighb anti bias correction. The linal \ tep .

for each unit vf (\Utput Ii:\ int! ｜｜ｴＺＱｧｨＱｾ＠ and bia:.. as "t:ll

a'> the hidden units are also lixing ''eights and bia'> [15 j.

'} J ｌＮｴＱｲｮＱＱＱｾ＠ Vt!ctur (!1u1111i:atio11 t LI

'<Ji

1.VQ algorithm i:. a method for trainmg competiti\e

la)crs of the supervised (16J . Competiti\e la)er ''ill

automaticall} learn 10 dassi f)' the given input vector.

I he approach tal..en is b) ､｡ｾｳｩ＠ f) the input 't>Cllir

hased on proximit) distam:e of the input 'ector to

agent \ICCtor (euclidean distance method).

L VO network structure is a l\\\1-la)1.'r neural

net'' \lrl.. Ｑｾ＠ 1.:lm1pnseJ of mrwt ｬ｡ｾ＠ er :ind output la) er

Input la)cr Clllltainc, neurone; ｡ｾ＠ man) a-. dimen;illllal

input. l.)Utput la) er contains neurons ｡ｾ＠ man)

as

the

number of classes

r

In

These I\\ 0 la) t:r'> are

...onnecteJ b) linl.. ｢･ｴ｜｜ｬＮｾ･ｮ＠ each neuron that ha:- a

certain weight is called the agent vecto r.

LVQ algorithm is a method of artificial neural networ!.. hased on competition with a mechani'>m squared euclidean distance in picking winnt>rs agent 'ector to determine the category of the input vector. In L VQ nt:twork. the learning process unJenal..en are a

;upervised learning. "here pro' iJing input are accompanied by the expected output information. etwork alway!> directed to determine the most appropriate output unit with the target of input 'e<:tors

through shift position of agent vector. If the vectors of

training Jata are grouped toge1her with ngcnt 'c:ctor ''inner. then the agent vector is shifted come near to vecto r of training data h) the equation ·

I\' (n6'\<·) = II'

.

(olci) + ol.\'. - I\'

..

(old) ( ()

If vectors the training data are not grouped together with the agent \.ector winner. then the agent 'ector

shifted :ma) from the training vector is ｣Ｚ＼ｰｲ･ｳｾ｣ｊ＠ by

the equation :

Ir.. ＨＱＱｾ ﾷ Ｉ＠

=

ii'.. (olo) - c(.'( - ［ｾﾷ Ｎ Ｈｯｬ､Ｉ＠ (::!)

Where :

0 JI'

r

I earning 1.11e

\\ 1:1gh1 ur .1gcnl \l.'..:lllr

｜ｾ･ｩｧｨｴ＠ \.l f lllput \C\.!Or

111 RI "' I t \ ｜ｾｬｬ＠ Dt\tl l,\IU'I

T <:sting JI ｴｨｩ ｾ＠ st:ige to lool.. at the lc'd of

rel.<'gn111on ｡ＮＮＺ｣ｵｲ｡｣ｾ＠ u<.ing 1.VQ Te)!lllg 1s done ｢ｾ＠

1.:hanging th .. · numher ,,f coelfo:u:nts :'\1fCC and tho!

\ alut: uf karn1ng r,Hc 10 Sl'C ii:, effect Oil the lc\el Of

rc..:ogrnlllllt accurat:) The number of ｾｉｆｃｃ＠

Ｎ［Ｌ Ｑ ＮＺＰｩ｣ｩ･ＱＱＱｾ＠ ｵ ｾ ･､＠ Jr.: MFCC 13. MFCC I'. MFCC 20

dan \1H'C Jll. Ｂｨ｣ｲ･ＮＱｾ＠ tht: learning rate used are 0.1, 0.2. O .. ;, 0 ｾｊ｡ｮ＠ 0 ｾ＠

I V<.>

1111p lcm.:ntallun ts done 111 Matlab

fum;tion Learning

2. I 1hat on Matlab

en' lfl>lllllClll \\ ith ; r .- '

ｨＱＱＱ｣ｴｩＱｬｬｬｾ＠ ｵｾＱＺｊ＠ art: I. VO

unplcmentcJ ''ith functiom :<.!c. u .lvq_.

A<:1:onJi11g 10 e'periment on L VQ ｢ｾ＠ vaf) ing the

purameter value' the learning rate :rnd cocdfo:ients

M ITC. obtained a1:eurac:v leH!ls \al) in testing \\Ith

tl:!:-.t data a., presented m Table I. There was no general

pallern of the: rdation:.hip between MFCC codlicit:nl

and learning rate. hut both paramct.:rl> affect ｴｨｾ＠

｡｣｣ｵｲ｡｣ｾ＠ frt:cl). I lo\\CH!r. the value of the highe:>t

accuracy in I. VQ obtained with 20 MFCC coefficients

and the of 0.2 lc:arning rate with an accurnc) 97.02°. {mar\..eJ in bold).

l.\Ht t-I

.\( < l RIC' tll' t FST"l1 I!\ I.VI}

\lln Lcarninl? Kat.:

( 1>.:ll'ic1..:11h

tit ti'.! 0 J 0 4

I ' 911 I!! I},\ Ｔｾ＠ 1)4 011 Ill -1-1

15 l(IPN 9).45 91 37 Jll 85

w

89 :W 97.U:? 76 19 8-182

Ｎ ｾｯ＠ l)l .26 8601 ll-182 'Xl-11

Overall. testing with LVQ can achie\e the highest

;iccurac) of I00° o. Ho\\ e-..er. in a ｭｩｮｯｲｩｾ＠ of

cxpcrimems obtained an accuracy of 0°ro, ie no data is

correctly identified . This is due to the performance of

the L VQ training up and ､ｯＱｾｮ＠ during the training

ｰｲｾＧ｣･ｳｳ＠ iternt ion or not towards convergent. For ･ｾ｡ｭｰｬ･Ｎ＠ it was determined that the maximum error of

0.0 I. howc' i:r \\hen th.: training process until the

ｭｮｾｩｭｵｭ＠ 11cra1io11 nc\er been obia111ed the error rate.

ve11 the error becomes higher as the iteration goes though the prt!\ ious iteration error oflow value

fl. Tes1111g Resu/11 o/ Backpropagation .\lethod

Testing al thi" stage to look at the level of recognition accuracy using Backpropagation. Testing

is done by changing the number of coefficients MFCC

ancJ the value of Jcaming rate to see its effect on the

'

'

ICACSIS 201 -1

k\cl ol n:cug.11i11un accura.,;) I he m1111hl.'r ,,f \ll·l·l

cocllirn:nl!'> used arc MFCC J l. ｾＱＱ＠ lT ｉｾＮ＠ t\1FCC 20

Jan ｾＱｊＭｃＮＺｃ＠ 30. \\hcn:a:- lhc learning 1a11: ｵｾＮＮＺｊ＠ arc 0 I.

0.2. 0.3. 0.4 dan 0.5.

fl11dpropagatiu11 impkmc:ntall\lll Ｑｾ＠ J,111c in Matlah

Cll\ironmcnt \1i1h nc,,ff Actl\alh'll fu11d11111 ｵｾ｣ｊ＠ Ｑｾ＠

\1gmoiJ 1 J _; '. ...i J in thl.' h 1JJc:11 l.1: l'r anJ a linc:ir

funct1llll ＨＺＮＮＡｾ ﾷＬＮＺＺＮ Ｉｩｮ＠ tht' ,1l1tput IJ:cr lrJ111ing

.1lgon1hm u,cJ ｩｾ＠ graJ1t'nt J.:,cc:nt 1 · : , " · :J.

Ac.:lirding Ill experiment on lhH:l..prupag:.11io11 h: \:tr) mg the: parameter 'ah1e\ the learning rate: anJ

coefficients MFCC. obtaineJ accurm:) k\ds 'ar: in

te5ting \\ilh ICSl data as prc:scmcd in fuble 2. ｊｵｾｴ＠ JiJ..i:

the L VQ. not obtaineJ u di:ar general paltc:rn of th<.' relationship between MFCC coefficient anJ karning r:tle. However. from lhe table! ii 1.:an he seen that the

｡｣｣ｵｲ｡｣ｾ＠ tenJs to increase '' i1h i111.:n:asi11g learning rate wi1h an increase in the: lluctuat1ng In some cases. the value of learning rate that is too high c,111 maJ..e ,1 lo" er accurac).

the value of the: ｨＱｧｨ｣ ｾ ｴ＠ accurac: 111

Backpropagation obtained "1th 11 M FCC

codlicient!'-.ind the of 0.4 learning rate with an accurac: 97 . .12" o

(fable 2). 11ad:propagati<'ll mcthml ab\1 1.::.111 ad1ic1 e

the highest accurac) of I 00° o a) in L VQ. hut in some

expcrinwnts. the accurac} is quill! lo\\ on :i particular

class can ewn reach 0° o.

r1\UI F II

,\cn, kACY OF 1 EST NG IN R \O..PR(ll' \(1-\ 110'1

\ll·CC: I .::irn1nti R:it.:

ro.:tlic1.:nb

0 I () 2 (I l 11 . .i fl 5

IJ 9t% %-13 C).j Ｈｉｾ＠ 97.32 96 7)

15 89 88 9.t l).j o.i ｯｾ＠ 1.J:'SJ 1)1) .jJ

20 87 5 9t 96 93 75 95 ).j lj.j .35

30 8036 89 58 115 ;!.j 'II Ill> t).jC).j

C Ejj'i?<:t of Numher t/ .\IFC(' C:oe//ic1t•111 Agmmt

r1cc urUC.J'

From 1he entire experiment. the a\\:rnge ｡｣｣ｵｲ｡｣ｾ＠

[image:24.619.266.508.177.474.2]

obmincd at each value of tht: coefticient is shO\\n in

Figure 5. From the graph it can be seen that there is a

lendency that the higher the coefficient M FCC makes

lo"er accuracy in L VQ and Backpropagation. ' although the decrease was not significant. Accuraq at

MFCC coefficient 13 tend to be higher than the

accuracy of the MFCC coenicicnt greater. This is due lo the higher coefficient of MFCC impact on the dimensions of a larger data. Large data dimc:nsions

make the generalization capability of /\NN is lower so

lhe accuracy is decreased.

D. Effect of Learning Rate Ag11111H ..lcrnrac:1·

Influence of the learning rate throughout the cxperimenl to the average accurac)' is shown in Figure

6. From the graph, in L VQ there is a 1enden1:) that

h1ghc,1 ＮＱＮＺｬｵｲＺＱＱＮＺｾ＠ lll·cur un the lcarn111g 1.1te '' 11\lt tl1l1

I''"

or h ll' high If the learning ratl' 1:-; ll'tl high. ＮＱ｣ｲｮｲ｡｣ｾ＠ 11 ill Ot.' ｊ･｣ｲ･｡ｾ｣＠ because of th.: L \'<) that Ull the: change l>f nct\\ork \1cights "1th tht.' nature of the 1:1 impl't it il111. 1c. increa.;e or ,k.:rc,1'l' the '' l' 1ght 111 thl· net\\11rJ... '\' that the karn111g 1.1ti: 1' 1,1,1 l.1rg,· .,;,11"ing Jr.1 ... t11 .. 1.hang.::. .11 l'ach 11era1i11n w th.it the \1eigh1 ut

tho: 11..:t\1 \1 rJ.. ｢･｣ｯｭ･ｾ＠ ｵＱＱｾｴＮＱｨｨＺ＠ I ll111 I!\ l!r. 1f the

li:arn1ng rate i-. t1l0 )011. thc lc;ir11111g. ｰｮＱ｣ｩＺＡ＾ｾ＠ ,,j)) he

h111g

l

..

!

c

•

•

...

I

"

..

.

.

l

...

..

....

..

...

.

ｾ＠ I

.

ﾷｾ＠ I

,

.

..

,

.. ｊｾ Ａ＠

...

..

: I,• ｾｉ＠

....

,

I)

)':'\°l;

\111 ....

...

, ..

ｾﾷﾷ ＾＠ .,.

I> ｾＮｬ＠

"''li1 ,,.,,:.1

...

,

,, ｾ Ｇ｜ｉｩＬ＠

.: I

..

,,,. .

.

"'

G.-.ph of accuracy Butd on Ltrning Rolt

)(<)

ﾷｾ＠

ﾷｾ＠

J

!:

'•4

ｾ＠

!':

..

ｾ＠ .. l

Ill

'.l!

c

!16

&.I

81

l!l)

tl! 0,/

.,.

Ill

ﾷ ﾷ ｾ ﾷｬｬ ＧＧ＠

...

t il.•• .. } 'l\. \) 'l\.'t\ ｾ ｉ＠ Ｇｾ＠ A'· 'It .

• ' ·11 • ( T"""' '"A o.n .. ) ＧｋｬＮＧｾ＠

.,

_ﾷｾ＠ flt• •e S\;t,

.p.,: r·,,.." .... ., o,.•.ai S?"l 91.'' ＧｬｾＧ＾＠ 'lh.G•

• RP t1 f\h r._ £> •'..ti ｾＬＮＬ＠ _{OJ,/ I 94,4/}

'''·

,,,

...

Ml S1

"'' ....

. .

""'

.. ) . .1/

j

tl.\ '1\1.t•

ｾ Ｎｩ ﾷ＠

} / , l j

ｾＧＭｾＧ＠

. . - - • ,p

hgurc Ii Graph <>f ｡｣ＮＺｵｲｮ｣ｾ＠ !3a(cd 11n f t<1111in.f! H111 ..

In contrast, in Backpropagation. increased learning rate tends to increase the accuracy. rhis is due to the

algorithm BacJ..propagation ｪｵｾｴ＠ ch;inging the nctworJ..

\\Cight-: ｢ｾ＠ -;umrning