Enginee ing and Technology Jou nal e-ISSN: 2456-3358
Volume 10 Issue 11 No embe -2025, Page No.-7923-7930
DOI: 10.47191/e j/ 10i11.23, I.F. – 8.482
© 2025, ETJ
7923
ETJ Volume 10 Issue 11 No embe 2025,
1
I an Dwi P awi a
Clus e ing o Sc ew P ess Machine Condi ions using he Agglome a i e
Hie a chical Clus e ing
I an Dwi P awi a1, Jas il2, Suwan o Sanjaya3, Les a i Handayani4, Fi i Insani5
1,2,3,4,5In o ma ics Enginee ing, Facul y o Science and Technology, Sul an Sya i Kasim Riau S a e Islamic Uni e si y.
ABSTRACT: The sc ew p ess machine is an impo an componen in he palm oil p ocessing p ocess ha unc ions o ex ac oil
om palm ke nels. Con inuous use o he machine can educe pe o mance and dis up he p oduc ion p ocess. The e o e, machine
condi ion analysis is needed o suppo p e en i e main enance s a egies. One me hod ha can be used is he clus e ing echnique.
Clus e ing is a echnique o g ouping da a based on speci ic pa ame e s o o m classes wi h simila cha ac e is ics. This s udy
applied he Agglome a i e Hie a chical Clus e ing (AHC) me hod wi h a single linkage app oach o g oup he condi ions o sc ew
p ess machines based on da a ob ained om PT. XYZ o he pe iod Ap il-May 2024, wi h a o al o 23,002 da a poin s. The esea ch
s ages included da a selec ion, da a p e-p ocessing, no maliza ion using Z-sco e, clus e ing wi h AHC, and e alua ion using he
Silhoue e Coe icien and Da ies-Bouldin Index (DBI). The esul s showed ha he AHC me hod was able o o m a ep esen a i e
g ouping o machine condi ions. E alua ion using he Silhoue e Coe icien p oduced he bes numbe o clus e s a 2 clus e s wi h
a alue o 0.591, indica ing ha he clus e ing quali y was in he good ca ego y. Meanwhile, e alua ion using DBI showed he bes
numbe o clus e s a 4 clus e s wi h a alue o 0.404, indica ing ha he sepa a ion be ween clus e s was qui e good. These indings
can be used as a e e ence in de e mining p e en i e machine main enance policies so as o inc ease p oduc ion e iciency.
KEYWORDS: Sc ew P ess, Agglome a i e Hie a chical Clus e ing, Clus e ing, Silhoue e Coe icien , Da ies-Bouldin Index
I. INTRODUCTION
In he palm oil indus y, F esh F ui Bunches (FFB) a e
p ocessed o p oduce c ude palm oil (CPO) and palm ke nel.
The success o his p ocess is la gely de e mined by he
pe o mance o p oduc ion machine y, one o which is he
sc ew p ess machine ha unc ions as a key componen in he
oil ex ac ion s age om palm ui [1]. Con inuous ope a ion
o he machine o e a long pe iod o ime can cause a decline
in pe o mance du ing ce ain pe iods. This condi ion no
only has he po en ial o hampe he smoo h unning o he
p oduc ion p ocesses, bu also inc eases he isk o wo kplace
acciden s and causes signi ican losses o he company [2].
The e o e, i is impo an o ensu e ha p oduc ion machines
a e always in op imal condi ion h ough he implemen a ion
o an e ec i e main enance sys em. Rou ine and pe iodic
main enance is a p e en i e measu e o main ain s able
machine pe o mance and suppo p oduc ion con inui y [3].
Dis up ion o he mechanical componen s o he sc ew p ess
machine will hampe he nex p ocessing s age and
au oma ically cause losses [4].
A sc ew p ess machine is a ool ha unc ions o con inue he
p ocess o sepa a ing oil om he diges e . This machine is
equipped wi h a double sc ew ha pushes he p essed mass
ou , while opposi e p essu e is applied h ough a hyd aulic
double cone. A his s age, he ui pulp ha has been s i ed
will be squeezed so ha he oil con ained in i can come ou
due o he p essing p essu e. In a s udy wi h a clus e ing case
s udy on sc ew p ess machines using he Fuzzy C-Means
algo i hm wi h e alua ion using he Elbow me hod and
Da ies-Bouldin Index (DBI), i was ound ha he Elbow
me hod p oduced ou op imal clus e s, while he DBI
p oduced wo op imal clus e s [5]. Ano he s udy wi h a
clus e ing case s udy on sc ew p ess machines using he K-
Means algo i hm wi h e alua ion using he Elbow me hod
and Da ies-Bouldin Index (DBI) showed ha he Elbow
me hod p oduced ou op imal clus e s, while he DBI
me hod p oduced h ee op imal clus e s [6]. The sc ew p ess
machine unc ions o squeeze chopped chips, c ushed om
he diges e , o p oduce c ude oil [7]. To imp o e oil
sepa a ion e iciency, he p essings om he sc ew p ess a e
lowed h ough a Back P essu e Vessel (BPV), which
unc ions o egula e back p essu e so ha oil ex ac ion is
maximized. The Back P essu e Vessel (BPV) is a p essu ized
essel used o collec and dis ibu e low-p essu e s eam o
p ocessing uni s in palm oil mills ha u ilize s eam as a powe
sou ce o equipmen [8]. The main ole o he BPV is o
main ain a con inuous supply o s eam o a ious p ocesses
in palm oil mills. The s eam is p oduced by boile s h ough
he p ocess o hea ing p essu ized wa e , hen pa o i is used
by u bines o gene a e elec ical ene gy, while he emaining
s eam is channeled o he BPV. F om he BPV, he s eam is
dis ibu ed o he s a ions ha need i . Thus, he s eam
capaci y p oduced by he boile mus be able o accommoda e
he needs o he en i e p ocessing chain in he palm oil mill
“Clus e ing o Sc ew P ess Machine Condi ions using he Agglome a i e Hie a chical Clus e ing”
7924
ETJ Volume 10 Issue 11 No embe 2025,
1
I an Dwi P awi a
[9]. Gi en he c ucial ole o sc ew p ess machines and Back
P essu e Vessels (BPV) in main aining he smoo h ope a ion
o he p oduc ion p ocess a palm oil mills, analyzing he
ope a ional da a o hese machines is essen ial. One app oach
ha can be used o g oup his da a is he clus e ing me hod.
Clus e ing is a me hod in da a mining ha alls unde he
unsupe ised ca ego y. In gene al, clus e ing me hods can be
di ided in o wo ypes, namely hie a chical clus e ing and
non-hie a chical clus e ing, bo h o which a e widely used in
he da a g ouping p ocess [10]. Clus e ing i sel can be
unde s ood as he p ocess o g ouping da a o objec s based
on he in o ma ion con ained in he da a, which ep esen s he
cha ac e is ics o objec s and hei ela ionships [11]. The
pu pose o his p ocess is so ha objec s included in one g oup
ha e high simila i y o ele ance, while objec s in di e en
g oups do no ha e such simila i y o ele ance [12]. The
pu pose o clus e ing is o iden i y o simpli y da a by
di iding i in o se e al small g oups so ha i is easie o
analyze. This p ocess is o en e e ed o as da a segmen a ion
[13]. Clus e ing can also be used o educe complex da a by
g ouping simila da a, he eby making he da a sho e
wi hou losing i s con en [14]. Based on he a ious
clus e ing me hods a ailable, he hie a chical app oach is
conside ed supe io o ce ain cases because i is able o
desc ibe he ela ionship be ween da a in he o m o a
hie a chical ee. One o he mos commonly used me hods is
Agglome a i e Hie a chical Clus e ing (AHC).
Agglome a i e Hie a chical Clus e ing (AHC) is a clus e ing
me hod ha uses da a analysis explo a ion echniques by
g ouping da a in o se e al g oups called clus e s [15]. AHC
is a da a g ouping me hod based on he le el o simila i y
be ween objec s, esul ing in a ep esen a ion in he o m o a
ee-like hie a chical s uc u e [16]. AHC uses se e al
app oaches o de e mine he dis ance be ween clus e s,
including single linkage, comple e linkage, a e age linkage,
and Wa d [17].
In his s udy, he au ho used he Agglome a i e Hie a chical
Clus e ing (AHC) me hod, which has ad an ages o e o he
clus e ing me hods because i does no equi e de e mining
he numbe o clus e s a he beginning o he p ocess. In
addi ion, his s udy applied a single linkage app oach, which
is a me hod ha de e mines he dis ance be ween clus e s
based on he closes p oximi y be ween objec s in each clus e
[18]. In a s udy wi h a case s udy o clus e ing o de e mine
s uden majo s using he AHC me hod, ou clus e s we e
p oduced, namely clus e 0 wi h 93 da a poin s, clus e 1 wi h
10 da a poin s, clus e 2 wi h 10 da a poin s, and clus e 3 wi h
8 da a poin s [19]. Ano he s udy wi h a case s udy o s uden
li e a u e clus e ing using he AHC me hod wi h a single
linkage app oach p oduced h ee clus e s, namely clus e 1
wi h 1840 da a, clus e 2 wi h 34 da a, and clus e 3 wi h 16
da a, wi h a o al o 1890 da a [20].
Based on he abo e explana ion and issues, he pu pose o his
s udy is o ob ain he bes numbe o clus e s using he
Agglome a i e Hie a chical Clus e ing me hod wi h o iginal
da a om PT.XYZ om Ap il 2024 o May 2024.
II. RESEARCH METHODOLOGY
A. Resea ch S ages
This esea ch consis s o se e al s ages ha he au ho ca ied
ou sys ema ically h ough a p ocess o analysis and
p ocessing o ele an da a. The i s s age was da a selec ion
o de e mine which da a would be used. Nex , p e-p ocessing
was ca ied ou by cleaning he da a so ha i was eady o
ans o ma ion. A e ha , he da a was ans o med, which
included a no maliza ion p ocess. The nex s age was
clus e ing using he Agglome a i e Hie a chical Clus e ing
(AHC) algo i hm. The inal s age was an e alua ion o
de e mine he bes numbe o clus e s p oduced by applying
he algo i hm. The esea ch low can be seen in Figu e 1.
Figu e 1. Resea ch Me hodology Flow
B. Da a Selec ion
A his s age, da a selec ion is ca ied ou o de e mine which
da a will be p ocessed so ha i is in line wi h he esea ch
equi emen s. The a iables and ypes o da a used a e
adjus ed o he esea ch objec i es so ha he selec ed da a
a e uly ele an . The pu pose o his s age is o ensu e ha
he da a is eady o analysis so ha i p oduces mo e accu a e
esul s.
C. Da a P e-P ocessing
One o he s ages in da a p e-p ocessing is da a cleaning. This
s age aims o ensu e ha he selec ed and combined da a a e
app op ia e and eady o use in he p ocessing s age. Da a
cleaning includes co ec ing in alid da a, emo ing i ele an
a ibu es, and checking and handling missing alues.
D. Da a T ans o ma ion
The ans o ma ion s age aims o ensu e ha he da a used can
be op imally p ocessed by he Agglome a i e Hie a chical
Clus e ing algo i hm. A his s age, da a no maliza ion is
ca ied ou o main ain he alidi y and consis ency o he da a
quali y. No maliza ion is he p ocess o ans o ming da a
in o a mo e uni o m, o ganized, and simple o m o analysis,
as well as educing inconsis encies be ween alues. This
s udy uses he Z-sco e me hod. Z-sco e con e s each da a
alue in o a s anda d dis ibu ion wi h a mean o 0 and a
s anda d de ia ion o 1, so ha da a om a ious a iables
a e on a compa able scale [21]. The ollowing is he o mula
o using he Z-sco e.
𝑍= 𝑋− µ
𝜎
(1)
“Clus e ing o Sc ew P ess Machine Condi ions using he Agglome a i e Hie a chical Clus e ing”
7925
ETJ Volume 10 Issue 11 No embe 2025,
1
I an Dwi P awi a
Whe e 𝑋 is he o iginal da a alue, µ is he mean alue, and
𝜎
is he s anda d de ia ion. By using his me hod, he da a
becomes mo e uni o m on a compa able scale, so ha
di e ences in alues be ween a ibu es do no domina e he
dis ance calcula ions used in clus e ing.
E. Applica ion o he Agglome a i e Hie a chical
Clus e ing Algo i hm
Agglome a i e Hie a chical Clus e ing (AHC) is a clus e ing
echnique ha s a s by ea ing each da a poin as i s own
clus e . The algo i hm hen i e a i ely me ges he clus e s ha
ha e he smalles dis ance be ween hem, o ming a
hie a chical s uc u e. One common me hod used in AHC is
Single Linkage (Nea es Neighbo ). In his app oach, he
dis ance be ween wo clus e s is de e mined based on he
minimum dis ance be ween pai s o objec s om di e en
clus e s. This app oach is sui able when he pu pose o he
analysis is o g oup da a based on he closes p oximi y
be ween objec s [22]. The applica ion o he Agglome a i e
Hie a chical Clus e ing algo i hm wi h he Single Linkage
app oach is as ollows [23].
a. Calcula e he dis ance be ween objec s using Euclidean
Dis ance :
𝑑𝑖𝑚= √∑(𝑥𝑖𝑗− 𝑥𝑚𝑗)2
𝑝
𝑗=1
(2)
Explana ion :
𝑑𝑖𝑚 = dis ance be ween objec i and objec m
𝑥𝑖𝑗 = alue o objec i in a iable j
𝑥𝑚𝑗 = alue o objec m in a iable j
𝑝 = numbe o a iables
Then o m he ma ix 𝐷={𝑑𝑖𝑚}.
b. De e mine he smalles dis ance in he ma ix 𝐷 and
combine he wo objec s in o one clus e .
c. Upda e he dis ance ma ix using he Single Linkage
app oach :
𝑑𝑈𝑉=min{𝑑𝑈𝑊,𝑑𝑉𝑊}
(3)
Explana ion :
𝑑𝑈𝑊 = dis ance be ween clus e U and clus e W
𝑑𝑉𝑊 = dis ance be ween clus e V and clus e W
d. Con inue un il all objec s a e g ouped in o a single
clus e .
F. E alua ion
In his s udy, e alua ion was pe o med using he Silhoue e
Coe icien and Da ies-Bouldin Index me hods.
a. Silhoue e Coe icien
The Silhoue e Coe icien is one o he e alua ion
me hods used o assess he quali y o clus e ing esul s.
This me hod measu es how well an objec is placed in
he app op ia e clus e compa ed o o he clus e s. The
close he Silhoue e Coe icien alue is o 1, he be e
he clus e ing quali y wi hin a clus e . Con e sely, i he
alue is close o -1, i indica es ha he clus e ing
quali y in ha clus e is poo o less han op imal [24].
𝑠(𝑖)= 𝑏(𝑖)−𝑎(𝑖)
max {𝑎(𝑖),𝑏(𝑖)}
(4)
Explana ion :
𝑎(𝑖) = a e age dis ance be ween objec 𝑖 and all
o he objec s in he same clus e
𝑏(𝑖) = a e age dis ance be ween objec 𝑖 and all
objec s in he nea es di e en clus e
𝑠(𝑖) = Silhoue e Coe icien alue o objec 𝑖
b. Da ies-Bouldin Index
The Da ies-Bouldin Index (DBI) is an e alua ion
me hod in clus e ing used o assess he quali y o
sepa a ion be ween clus e s. The measu emen is based
on he le el o cohesion wi hin a clus e and he le el o
sepa a ion be ween clus e s. A smalle DBI alue
indica es be e clus e ing quali y, as i indica es ha he
clus e s o med a e mo e compac and ha e a clea
sepa a ion dis ance be ween one ano he [25]. The
ollowing a e he s eps o calcula ing he Da ies-
Bouldin Index (DBI).
1. Calcula e he Sum o Squa es Wi hin-Clus e
(SSW) alue
𝑆𝑆𝑊𝑖= 1
𝑚𝑖∑𝑑(𝑥𝑗,𝑐𝑖)
𝑚𝑖
𝑗=𝑖
(5)
Explana ion :
𝑚𝑖 = numbe o da a poin s in he i- h
clus e 𝑖
𝑐𝑖 = cen oid o he clus e 𝑖
𝑑(𝑥𝑗, 𝑐𝑖) = dis ance o each da a poin o
cen oid i calcula ed using Euclidean dis ance
2. Calcula ing he Sum o Squa es Be ween-Clus e
(SSB) alue
The sum o squa es be ween clus e s (SSB) is used
o assess he le el o sepa a ion, which is calcula ed
by measu ing he dis ance be ween he clus e
cen oid and he o e all da a cen e . The highe he
SSB alue, he be e he sepa a ion be ween
clus e s. The ollowing is he SSB calcula ion
o mula.
𝑆𝑆𝐵𝑖=(𝑐,𝑐𝑗)
(6)
Whe e, 𝑑(𝑥𝑖,𝑥𝑗) is he dis ance be ween da a poin 𝑖
and da a poin 𝑗 in ano he clus e .
3. Calcula e he a io o he Sum o Squa e Wi hin-
Clus e (SSW) alue and he Sum o Squa e
Be ween-Clus e (SSB)
A good clus e is one ha has he smalles possible
cohesion alue and he la ges possible sepa a ion
alue. This a io is used o compa e clus e s,
pa icula ly be ween he𝑖 and𝑗 clus e s, whe e he
indices 𝑖 ,𝑗 , and 𝑘 ep esen he o al numbe o
clus e s o med.
𝑅𝑖𝑗= 𝑆𝑆𝑊𝑖+ 𝑆𝑆𝑊𝑖
𝑆𝑆𝐵𝑖𝑗
(7)
Explana ion :
“Clus e ing o Sc ew P ess Machine Condi ions using he Agglome a i e Hie a chical Clus e ing”
7926
ETJ Volume 10 Issue 11 No embe 2025,
1
I an Dwi P awi a
𝑆𝑆𝑊𝑖 = Sum o Squa e Wi hin-Clus e a he
cen oid 𝑖
𝑆𝑆𝐵𝑖𝑗 = Sum o Squa e Be ween-Clus e da a
be ween he𝑖 and𝑗 in di e en clus e s
4. Calcula e he Sum o Squa e Be ween-Clus e
(DBI) alue om he Ra io alue ob ained
p e iously
𝐷𝐵𝐼= 1𝑘∑𝑚𝑎𝑥𝑖≠𝑗(𝑅𝑖𝑗)
𝑘
𝑖=1
(8)
Explana ion :
Whe e 𝑘 is he numbe o clus e s used. The smalle
he Da ies-Bouldin Index (DBI) alue ob ained
(non-nega i e, ≥ 0), he be e he quali y o he
clus e s o med. Con e sely, i he DBI alue
ob ained is qui e la ge, his indica es ha he
clus e ing esul s using he clus e ing algo i hm a e
subop imal.
III. RESULT AND DISCUSSION
A. Da a Selec ion
This s udy uses da a ob ained di ec ly om PT.XYZ is a
company engaged in palm oil p ocessing. This s udy ocuses
on sc ew p ess machines, which play an impo an ole in he
palm oil ex ac ion p ocess. The da a used comes om a
sc ew p ess machine wi h he de ice code BPV EP01-09,
which se es as a unique iden i ie o he machine in he
company's moni o ing sys em. A o al o 23,002 da a poin s
we e collec ed, co e ing a ious ope a ional pa ame e s such
as machine code, p essu e, and empe a u e. A e he inpu
p ocess, only wo a ibu es we e selec ed o use, namely
empe a u e and p essu e, as hese wo a ibu es would be
used o u he p ocessing. The ollowing a e he esul s o
he da a a e he selec ed da a was ex ac ed, as shown in
Table 1.
Table 1. Selec ed Da a
NO
TEMPERATURE
PRESSURE
0
123
3,19
1
123
3
2
123
2,95
3
123
2,77
4
123
2,83
...
...
...
22997
0
2,99
22998
100
1,62
22999
100
1,53
23000
100
2,12
23001
100
2,03
A e going h ough he selec ion s age, he da a used in his
s udy a e p esen ed in Table 1. The ocus o he s udy was
di ec ed a wo selec ed a ibu es, as hey we e conside ed
he mos ele an and had a signi ican in luence. The
selec ion o a ibu es conside s da a quali y, con ibu ion o
a ge a iables, and sui abili y o he con ex o he p oblem
being s udied.
B. Da a P e-P ocessing
This s age is a c ucial s ep o ensu e he quali y and sui abili y
o da a be o e i is used in he analysis and modeling p ocess.
In his s udy, p e-p ocessing ocuses on da a cleaning, which
aims o emo e o co ec inconsis en da a, missing alues,
and in alid da a. The da a cleaning p ocess is ca ied ou o
imp o e da a quali y and make i mo e e ec i e o use in he
nex s age o analysis. The da a used was sou ced om
PT.XYZ in aw o m. A e selec ion based on de ice codes,
he nex s age was o check o possible duplica ions and
missing alues, as well as o ensu e ha he da a o be
analyzed was o a uni o m and ele an ype. Wi h his
checking and cleaning, he da ase became mo e eady and
e ec i e o use, esul ing in mo e accu a e clus e ing esul s.
The e a e wo a ibu es used, namely TEMPERATURE and
PRESSURE, each o which has he same amoun o da a,
namely 23,002 da a poin s, no missing alues (No -Null), and
a consis en da a ype, namely loa 64. These condi ions
indica e ha he da ase has unde gone a ho ough da a
cleaning p ocess and is eady o u he analysis. Wi h no
missing alues o da a ype di e ences, his clean da ase is
eady o be used in he da a ans o ma ion s age o imp o e
he e ec i eness o sc ew p ess machine condi ion modeling.
C. Da a T ans o ma ion
Da a ans o ma ion is necessa y o add ess he di e ence in
scale be ween he empe a u e (0–157) and p essu e (0– 4.43)
a ibu es. Wi hou no maliza ion, he a ibu e wi h he la ge
alue ange, namely empe a u e, will be mo e dominan in
he dis ance calcula ion p ocess in he Agglome a i e
Hie a chical Clus e ing (AHC) algo i hm wi h he Single
Linkage app oach. This s udy uses he Z-sco e me hod. This
ans o ma ion changes he alue o each a ibu e based on
he mean and s anda d de ia ion o ha a ibu e, so ha he
da a has a mean o 0 and a s anda d de ia ion o 1. The esul s
o he ans o med da a can be seen in Table 2.
Table 2. Da a A e T ans o ma ion
NO
TEMPERATURE
PRESSURE
0
0,693
0,827
1
0,693
0,589
2
0,693
0,526
3
0,693
0,301
4
0,693
0,376
...
...
...
22997
-1,935
0,576
22998
0,202
-1,139
22999
0,202
-1,252
23000
0,202
-0,513
23001
0,202
-0,626
“Clus e ing o Sc ew P ess Machine Condi ions using he Agglome a i e Hie a chical Clus e ing”
7927
ETJ Volume 10 Issue 11 No embe 2025,
1
I an Dwi P awi a
D. Applica ion o he Agglome a i e Hie a chical
Clus e ing Algo i hm
The clus e ing p ocess in his s udy was pe o med using he
Agglome a i e Hie a chical Clus e ing algo i hm on 23,002
da a poin s ha had unde gone no maliza ion and consis ed
o wo main a ibu es, namely TEMPERATURE and
PRESSURE. The pu pose o applying his me hod was o
o m se e al clus e s ha ep esen ed simila i ies in he
ope a ional pa e ns o sc ew p ess machines. A each
agg ega ion s age, he Agglome a i e Hie a chical Clus e ing
algo i hm was applied using o mulas (2) and (3) consis en ly
on he no malized da a (Table 2). In he clus e ing p ocess, a
h eshold alue is used o de e mine he maximum dis ance
be ween clus e s o be me ged. This h eshold alue se es as
a cu -o dis ance on he dend og am, which is he poin a
which he clus e me ging p ocess is s opped. Wi h a
h eshold, he sys em can au oma ically o m he numbe o
clus e s ha ma ches he exis ing da a s uc u e wi hou he
need o de e mine he numbe o clus e s a he ou se . The
selec ion o he app op ia e h eshold a ec s he inal
g ouping esul s. The smalle he h eshold alue, he mo e
clus e s a e o med, and con e sely, he la ge he h eshold
alue, he ewe clus e s a e p oduced. Th ough his
app oach, each da a poin ob ains a label acco ding o he
clus e o which i belongs, so ha he g ouping esul s a e
able o desc ibe he condi ion o he machine in a mo e
s uc u ed g oup. The in o ma ion om his clus e ing can
hen be used as a basis o u he analysis, such as iden i ying
abno mal condi ions o de ec ing po en ial machine damage.
The es esul s can be seen in Table 3.
Table 3. Tes Resul wi h 4 Clus e s
CLUSTER
NUMBER OF DATA
C0
18529
C1
4396
C2
69
C3
8
Based on Table 3, he esul s o he da a a e es ing using he
Agglome a i e Hie a chical Clus e ing algo i hm show he
numbe o da a in each clus e as shown in Table 5. The
numbe o da a in C0 is 18,529, he numbe o da a in C1 is
4,396, he numbe o da a in C2 is 69, and he numbe o da a
in C3 is 8. The nex s ep is isualiza ion. The isualiza ion
can be seen in Figu e 2.
Figu e 2. Visualiza ion o clus e ing esul s
E. E alua ion
A e applying he Agglome a i e Hie a chical Clus e ing
algo i hm wi h a ious clus e numbe s, he nex s ep is o
e alua e and in e p e he quali y o he clus e ing esul s. The
pu pose o his s ep is o assess he ex en o which he da a
has been success ully g ouped acco ding o i s cha ac e is ics,
while ensu ing ha he numbe o clus e s selec ed is uly
op imal. In his s udy, wo e alua ion me hods we e used,
namely he Silhoue e Coe icien and he Da ies-Bouldin
Index (DBI). These wo me hods complemen each o he in
de e mining he numbe o clus e s ha a e mos
ep esen a i e o he da a s uc u e, so ha he clus e ing
esul s ob ained a e no only echnically accu a e bu also
ele an in he con ex o analyzing he condi ion o sc ew
p ess machines.
a. Silhoue e Coe icien
E alua ing clus e ing esul s is e y impo an o ensu e
ha he di ision o da a in o clus e s is op imal. The
Silhoue e Coe icien me hod can be used o assess he
quali y o each clus e by looking a he esul ing
coe icien alue. The close he alue is o 1, he be e
he clus e ing quali y because i indica es ha he da a is
mo e simila o he clus e i belongs o han o o he
clus e s. Con e sely, a alue close o 0 indica es o e lap
be ween clus e s. The ollowing a e he esul s o he
Silhoue e Coe icien based on he numbe o clus e s
es ed. The esul s o he e alua ion using he silhoue e
coe icien can be seen in Table 4.
Table 4. E alua ion o Clus e Resul Using he Silhoue e
Coe icien
CLUSTER
SILHOUETTE VALUE
2
0,591
3
0,565
4
0,553
5
-0,176
6
-0,287
7
-0,153
8
-0,169
9
-0,183
10
-0,184
Based on Table 4, he clus e ing esul s we e e alua ed using
he Silhoue e Coe icien me hod o assess he quali y o da a
sepa a ion in each clus e . The close he alue is o 1, he
be e he clus e ing quali y because i indica es a clea e
dis ance be ween clus e s. The Silhoue e Coe icien was
calcula ed using o mula (4) on he da a shown in Table 2,
wi h he calcula ion esul s p esen ed in Table 4. Based on
hese esul s, he highes Silhoue e alue was ob ained o
wo clus e s (k = 2) wi h a sco e o 0.591. This indica es ha
he con igu a ion wi h ou clus e s p o ides he mos op imal
sepa a ion and densi y. To cla i y he e alua ion end, he
calcula ion esul s a e isualized in he o m o a dend og am
so ha he pa e n o changes in he Silhoue e alue agains
“Clus e ing o Sc ew P ess Machine Condi ions using he Agglome a i e Hie a chical Clus e ing”
7928
ETJ Volume 10 Issue 11 No embe 2025,
1
I an Dwi P awi a
a ia ions in he numbe o clus e s can be obse ed mo e
in ui i ely and suppo he de e mina ion o he mos
app op ia e numbe o clus e s. The ollowing is he
e alua ion esul wi h he Silhoue e Coe icien as shown in
Figu e 3.
Figu e 3. E alua ion G aph Using Silhoue e Coe icien
b. Da ies-Bouldin Index (DBI)
E alua ing clus e ing esul s plays an impo an ole in
ensu ing ha da a di ision in o clus e s is op imal. Using he
Da ies-Bouldin Index (DBI) me hod, he quali y o each
clus e can be measu ed by selec ing he numbe o clus e s
ha p oduce he smalles DBI alue. The lowe he DBI alue
ob ained, he be e he quali y o he g ouping. The ollowing
shows he esul s o DBI alue calcula ions based on he
numbe o clus e s es ed. The esul s o he e alua ion using
DBI can be seen in Table 5.
Table 5. E alua ion o Clus e Resul s Using DBI
CLUSTER
DBI VALUE
2
0,586
3
0,453
4
0,404
5
1,394
6
1,523
7
1,667
8
1,571
9
1,527
10
1,453
Based on Table 5, he e alua ion o clus e ing esul s using
he Da ies-Bouldin Index (DBI) me hod aims o assess he
quali y o da a di ision in o each clus e , whe e he smalle
he DBI alue ob ained, he be e he clus e ing quali y. The
DBI calcula ion was pe o med using o mulas (5) o (8) on
he da a in Table 2, esul ing in he DBI alues shown in Table
5. F om hese calcula ions, i can be seen ha he lowes DBI
alue was achie ed wi h h ee clus e s (k = 4) wi h a alue o
0.404. This indica es ha he con igu a ion wi h ou clus e s
p o ides he mos op imal sepa a ion and densi y. To p o ide
a clea e pic u e o he e alua ion esul s end, a isualiza ion
in he o m o a dend og am was also ca ied ou , so ha he
pa e n o changes in DBI alues agains a ia ions in he
numbe o clus e s can be obse ed mo e in ui i ely and
suppo decision-making ega ding he mos app op ia e
numbe o clus e s.
Figu e 4. E alua ion G aph Using DBI
F. Compa ison o he Bes Clus e Resul s
A compa ison o clus e ing esul s using he Silhoue e
Coe icien and DBI me hods was conduc ed o de e mine he
mos app op ia e numbe o clus e s, aking in o accoun he
cha ac e is ics o each clus e o med. The bes clus e esul s
ob ained om he Silhoue e Coe icien me hod we e 2
clus e s, while he bes clus e esul s om he DBI me hod
we e 4 clus e s. The ollowing a e he esul s o he 2-clus e
and 4-clus e es s.
Table 6. Tes ing o 2 clus e s
CLUS
TER
NUMBER
OF DATA
TEMPERA
TURE
PRESSURE
C0
4473
0
0,02-3,32
C1
18529
92,5-157
0-4,43
Table 7. Tes ing o 4 clus e s
CLUS
TER
NUMBER
OF DATA
AVERAGE
TEMPERA
TURE
AVERAGE
PRESSURE
C0
18529
92,5-157
0-4,43
C1
4396
0
0,77-3,32
C2
69
0
0,02-0,23
C3
8
0
0,44-0,61
Based on Table 6 and Table 7, he esul s o clus e ing
e alua ion using wo me hods, namely Silhoue e Coe icien
and Da ies-Bouldin Index (DBI), ob ained di e en esul s
ega ding he bes numbe o clus e s. The Silhoue e
Coe icien me hod indica es ha he op imal numbe o
clus e s is 2, because a his numbe he Silhoue e alue is
“Clus e ing o Sc ew P ess Machine Condi ions using he Agglome a i e Hie a chical Clus e ing”
7929
ETJ Volume 10 Issue 11 No embe 2025,
1
I an Dwi P awi a
highe , indica ing clea e da a sepa a ion, compac ness, and a
good le el o simila i y wi hin clus e s. Meanwhile, he
Da ies-Bouldin Index (DBI) me hod shows he bes esul s
wi h 4 clus e s, because he DBI alue p oduced is lowe ,
which means ha he le el o simila i y be ween clus e s is
smalle and he dis ance be ween clus e s is ela i ely be e .
In his s udy, he use o he Silhoue e Coe icien me hod
wi h he o ma ion o 2 clus e s is conside ed mo e
app op ia e because i is able o p o ide a mo e concise,
s uc u ed da a sepa a ion ha is easie o unde s and in he
analysis p ocess.
IV. CONCLUSION
This s udy discusses he applica ion o he Agglome a i e
Hie a chical Clus e ing (AHC) algo i hm wi h a single-
linkage app oach in g ouping sc ew p ess machine condi ions
based on wo main a ibu es, namely empe a u e and
p essu e. The da a used is o iginal da a om PT. XYZ o he
pe iod Ap il - May 2024 wi h a o al o 23,002 da a poin s.
The esea ch p ocess was ca ied ou in s ages h ough da a
selec ion, p e-p ocessing wi h da a cleaning, da a
ans o ma ion using Z-Sco e no maliza ion, applica ion o
he AHC algo i hm, and e alua ion o clus e ing esul s using
he Silhoue e Coe icien and Da ies-Bouldin Index (DBI)
me hods. The esul s showed ha he clus e ing p ocess
di ided he da a in o se e al g oups ha e lec ed he
condi ions o he sc ew p ess machine in di e en si ua ions.
Based on he e alua ion using he Silhoue e Coe icien , he
bes numbe o clus e s was ob ained in wo clus e s wi h a
alue o 0.591. This indica es ha he wo-clus e
con igu a ion is able o p o ide clea e , mo e compac , and
s uc u ed da a sepa a ion. Meanwhile, he e alua ion esul s
using he Da ies-Bouldin Index (DBI) showed ha he bes
numbe o clus e s was ou clus e s wi h a DBI alue o
0.404, which indica es ha he quali y o sepa a ion be ween
clus e s is qui e good. The di e ence in hese e alua ion
esul s indica es a di e ence in ocus be ween he isual
cla i y o clus e ing and he le el o sepa a ion be ween
clus e s. O e all, his s udy p o es ha he AHC algo i hm
wi h he Single Linkage app oach can be used o analyze he
condi ion o sc ew p ess machines in a mo e sys ema ic and
s uc u ed manne . These clus e ing esul s can be used as a
basis o de e mining he numbe o clus e s ha a e mos
ep esen a i e o da a pa e ns. The limi a ion o his s udy is
ha i only uses wo main a ibu es, so o u he esea ch,
i is ecommended o add o he a ibu es in o de o ob ain
mo e comp ehensi e and accu a e clus e ing esul s.
REFERENCES
1. Y. Ningsih, M. Khuda i, T. Linangsa i, M. Zein, and
E. Les a i, “E aluasi Kine ja Mesin Sc ew P ess
Melalui Pene apan To al P oduc i e Main enance
(TPM) di Pab ik Kelapa Sawi PT. XYZ,” J. Teknol.
Ag o-Indus i, ol. 12, no. 1, pp. 61–69, 2025,
doi: 10.34128/j ai. 12i1.234.
2. M. L. Volodymy Ha an, “De e mina ion o
Hoppe Fullness o Sma Sc ew P ess Using
Machine Lea ning,” Compu . Des. Sys . Theo y
P ac ., ol. 6, no. 1, pp. 161–168, 2024,
doi: 10.23939/cds2024.01.161.
3. F. Pohan, I. Sapu a, and R. Tua, “Scheduling
P e en i e Main enance o De e mine Main enance
Ac ions on Sc ew P ess Machine,” J. Ris. Ilmu Tek.,
ol. 1, no. 1, pp. 1–12, 2023,
doi: 10.59976/ju i . 1i1.4.
4. D. Wa dian o and An inal, “Analisis kegagalan
mesin sc ew p ess ailu e analysis o he sc ew p ess
machine,” J. Tek. Mesin Insi u Teknol. Padang, ol.
12, no. 1, pp. 2089–4880, 2022.
5. J. Jas il, M. F. Al Fiq i, S. Sanjaya, L. Handayani,
and F. Insani, “Pengelompokan Da a Kondisi Mesin
Sc ew P ess Menggunakan Algo i ma Fuzzy C-
Means,” In . Sys . J., ol. 8, no. 01, pp. 60–70, 2025,
doi: 10.24076/in osjou nal.2025 8i01.2133.
6. F. K. Rahman, J. S. Sanjaya, L. Handayani, and F.
Insani, “Pene apan Algo i ma K-Means Clus e ing
pada Kine ja Mesin Sc ew p ess,” Bull. In .
Technol., ol. 6, no. 2, pp. 59–70, 2025,
doi: 10.47065/bi . 5i2.1783.
7. M. I. Pasa ibu, D. A. A. Ri onga, and A. I wan,
“Analisis Pe awa an (Main enance) Mesin Sc ew
P ess Di Pab ik Kelapa Sawi Dengan Me ode
Failu e Mode and E ec Analysis (Fmea) Di P .
Xyz,” Ji ekh, ol. 9, no. 2, pp. 104–110, 2021,
doi: 10.35447/ji ekh. 9i2.432.
8. Z. E endi, I. U. P. Rangku i, and M. D. Nug oho,
“Analisa Penga uh Laju Massa Uap Ra a-Ra a
Te hadap Kuali as Uap Ra a-Ra a Pada Pipa BPV (
Back P essu e Vessel ) Ke S e ilize Di Pab ik
Kelapa Sawi Kapasi as 50 Ton / Jam Dalam
be kembangnya indus i minyak kelapa sawi yang
mengalami kemajuan sanga pe,” ol. 13, no. 01, pp.
148–155, 2024.
9. M. Wisnu, G. Sup iyan o, and He man o o,
“Analisis Kebu uhan Uap pada Pe ebusan Tiga
Puncak,” Ag o o e ech, ol. 1, pp. 685–692, 2023.
10. M. No shahlan, H. Jaya, and R. Kus ini, “Pene apan
Me ode Clus e ing Dengan Algo i ma K-means
Pada Pengelompokan Da a Calon Siswa Ba u,” J.
Sis . In . T iguna Dha ma (JURSI TGD), ol. 2, no.
6, p. 1042, 2023, doi: 10.53513/ju si. 2i6.9148.
11. S. D. K. Wa dani, A. S. A iyan o, M. Um oh, and D.
Rolliawa i, “Pe bandingan Hasil Me ode Clus e ing
K-Means, Db Scanne & Hie a chical Un uk
Analisa Segmen asi Pasa ,” JIKO (Ju nal In o m.
dan Kompu e ), ol. 7, no. 2, p. 191, 2023,
doi: 10.26798/jiko. 7i2.796.
12. R. A. Ananda, “Clus e ing Menggunakan Algo i ma
K-Means un uk Mengelompokan Da a Pe judian
Be dasa kan Wilayah di Ko a Binjai ( S udi Kasus :
“Clus e ing o Sc ew P ess Machine Condi ions using he Agglome a i e Hie a chical Clus e ing”
7930
ETJ Volume 10 Issue 11 No embe 2025,
1
I an Dwi P awi a
Pengadilan Nege i Binjai ),” no. 4, 2024.
13. S. Magh i oh, A. P. Wijaya, and A. Hidaya ,
“Implemen asi Clus e ing Pada Da a Me e ing
Mesin Di S asiun T ansmisi T ans T Sema ang,”
P os. Sains Nas. dan Teknol., ol. 13, no. 1, p. 161,
2023, doi: 10.36499/psns . 13i1.9078.
14. P. Tang, H. Tang, W. Wang, and Y. Liu,
“Con as i e Clus e ing,” Lec . No es Compu . Sci.
(including Subse . Lec . No es A i . In ell. Lec .
No es Bioin o ma ics), ol. 13833 LNCS, pp. 294–
305, 2023, doi: 10.1007/978-3-031-27077-2_23.
15. Ul a ul Syaha a, Es y Ku niawa i, Ma io Pu a
Suhana, Rika Angg aini, and Falmi Yand i,
“Pene apan Me ode AHC (Agglome a i e
Hie a chical Clus e ing) un uk Klasi ikasi Habi a
Ben ik di Desa pengudang, Kabupa en Bin an,”
INSOLOGI J. Sains dan Teknol., ol. 3, no. 3, pp.
306–314, 2024, doi: 10.55123/insologi. 3i3.3547.
16. K. P a ama Simanjun ak and U. Khai a, “Ho spo
Clus e ing in Jambi P o ince Using Agglome a i e
Hie a chical Clus e ing Algo i hm,” MALCOM
Indones. J. Mach. Lea n. Compu . Sci., ol. 1, no. 1,
pp. 7–16, 2021.
17. R. P. Jus i ia, N. Hidaya , and E. San oso,
“Implemen asi Me ode Agglome a i e Hie a chical
Clus e ing Pada Segmen asi Pelanggan Ba be shop
(S udi Kasus : RichDjoe Ba be shop Malang),” J.
Pengemb. Teknol. In . Dan Ilmu Kompu ., ol. 5, no.
3, pp. 1048–1054, 2021, [Online]. A ailable:
h ps://j-p iik.ub.ac.id/index.php/j-
p iik/a icle/ iew/8730
18. R. Kusumas u i, E. Bayunanda, A. M. Ri a’i, M. R.
G. Asga , F. I. Ilmawa i, and K. Kus ini, “Clus e ing
Ti ik Panas Menggunakan Algo i ma
Agglome a i e Hie a chical Clus e ing (AHC),”
CogITo Sma J., ol. 8, no. 2, pp. 501–513, 2022,
doi: 10.31154/cogi o. 8i2.438.501-513.
19. R. T. Aldisa, “Da a Mining Penen uan Ju usan
Siswa Menggunakan Me ode Agglome a i e
Hie a chical Clus e ing (AHC,” J. Media In o m.
Budida ma, ol. 7, no. 2, p. 873, 2023, doi:
10.30865/mib. 7i2.6092.
20. A. N. Fadhilah and A. Janan o, “Klas e isasi
Li e a u Mahasiswa Menggunakan Me ode Ahc Di
Dinas Kea sipan Dan Pe pus akaan P o insi Jawa
Tengah,” J. Din. In o m., ol. 13, no. 1, pp. 09–17,
2021, doi: 10.35315/in o ma ika. 13i1.8366.
21. S. E. Saqila, I. P. Fe ina, and A. Iskanda , “Analisis
Pe bandingan Kine ja Clus e ing Da a Mining
Un uk No malisasi Da ase ,” J. Sis . Kompu . dan
In o m., ol. 5, no. 2, p. 356, 2023,
doi: 10.30865/json. 5i2.6919.
22. A. P. Wijaya, A. S. Kinan hi, D. Yanawa i, and S.
Syamsida , “Pengelompokkan Kabupa en / Ko a di
Pulau Jawa Be dasa kan Fak o Kemiskinan
Menggunakan Me ode Hie a chical Clus e ing,” J.
Sains dan Manaj., ol. 13, no. 1, pp. 40–51, 2025.
23. C. B. G. Allo, “Clus e ing Kabupa en/Ko a di
P o insi Papua Be dasa kan P oduk Domes ik
Regional B u o Menu u Lapangan Usaha
Menggunakan Single Linkage dan K-Medoids,” J.
Gaussian, ol. 13, no. 1, pp. 111–120, 2024, doi:
10.14710/j.gauss.13.1.111-120.
24. T. Rahmawa i, Y. Wilanda i, and P. Ka ikasa i,
“Analisis Pe bandingan Silhoue e Coe icien Dan
Me ode Elbow Pada Pengelompokkan P o insi Di
Indonesia Be dasa kan Indika o Ipm Dengan K-
Medoids,” J. Gaussian, ol. 13, no. 1, pp. 13–24,
2024, doi: 10.14710/j.gauss.13.1.13-24.
25. S. G. Ahmad, D. A i ian o, and W. Suha so, “Ju nal
Sma Teknologi Ju nal Sma Teknologi,” Uni .
Muhammadiyah Jembe , ol. 3, no. 5, pp. 502–510,
2022.
