A minor project on

K - means Clustering with different iterations

DONE BY

SHAIK SOUKAT ALI

K MEANS CLUSTERING
 between B and 1 using our own function.
iris.new<- -ris[,c(l,2,’',3))
iris.class<- iris[,”5pecics"]
head(-r-s.new)

4i Sc al.Larry.|i /›epel.l'iidih Petal.Len:.|: Metal.'?Tidt[.

4i 1 5.. .,.? 1.4 0.2
fi 2 4.9 3.0 1.4 0.2
f# 3 4.7 3.2 1.3 0.2
#i 4 4.6 3.1 1.5 0.2
#i 5 5.0 3.6 1.4 0.2
#i 6 5.4 3.9 1.7 0.4

head (i r i s . c e s s ;

#i [1] setosa setosa setosa setosa setosa setosa

#i Levels: setosa versicolor virginica

normalize <- function(x)(

return ((x-min(x))/ (max(x7-min(x)))

ins.new??e)al.Lend-)i<- normalise(iris.frs'?Se:al.Ler:(;t)i)
iris.newS%epal.Width<- -ormalize(iris.rev Nepal.AiJth)
-r-s.newSFetal.Le.ng’.)i<- )oz-iiializu(uis.nc^.ñPctal.Lai.‹TLIi)
iris.newsPetal.Width<- nornoli e(iris.new$°e’al.Width)
head(iris.new,

## ePal.Lerij h Nepal.Width Pet-l.Leg-L Petal.Wi.dLk

ii 1 %.22222222 J.ñ2a01'30 J.1C7?%ñ6l U.Uñl6?*%?
fi 2 U.?ñbñ66h U.4i* b6i 0.UDP 9b')l U.Uñlhc6%
ii 3 0.11111111 C.tJUJU?U ?.Url°'.’-.5 U.Uñlh?*.%
ii 4 O.U."?J?33J C.4ñ6?3°?.0.°›'. '.h’5 U.Uñl5c°%
#i 5 0.1'444444 U.o6666c’.' U.Ub’.'796o U.U4I6c6b.'
fi 6 U.G3d53l5ñ U.’/Sl66c’.' 0.l18hJ407 U.l?iUU1UC

results- Kmeans( iis.new,d) --,::-'..’ -' ...:.

#i [1] 61 39 50

resultFr=nrers # o:res value of cluster center ds! :’po:.’:' value{: centers for <-.’

#i 5Cpel.LCng:h Sepal.Wid-h PcteI .Lend:b Petal.Wndtk

#i 1 0.%(12556 2.3G73770 O.ÜÜ5715'IE 3.?'1*l6H?*Ü
#i 2 ü. 0/26bC 0.4508547 0.79704476 0.62'/663?
#i 3 0.1961111 (J.ú9 JUJH H.U t:JH)?? 0.Ü5OE3i’3.?

result$cluster ^cives c7uster vector =.L. , 'g the o7uster wLere eaof record fa77s

5. Apply k-means clustering algorithm with k = 3

 -- [38] 3 3 3 3 3 3 3 3 3 3 3 3 3 2 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
-- [75] 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 2 2 2 2 1 2 2 2 2
.. [11?]? ? 1 ? ? ? 1 ? 1 ? 1 ? ? 1 1 ? ? ? ? 1 1 ? ? ? 1 ? ? ? 1 ? 1 ?

9. Plot to see how Sepal.Length and Sepal .Width data po›nts have Deen d›stributed in clusters
1.0
0.8

O O
SepaI.Width

O O O OO O O
O O
0.4

O0 O OO O O O O O

. O O O OO O O
O .0

0.0 0.2 0.4 0.6 0.8 1.0

Sepal. Length

10. Plot to see how Sepal.Length and Sepal.Width data po›nts have been distributed originally as per ’ class” attr›bute ›n dataset
 1.0
0.8

OO O
SepaI.Width

0.6

O 0 O O O OO O O
0 O O O O
O O0
0.4

Q O O O

OO O OO O O O O O
0.2

G O

0.0 0.2 0.4 0.6 0.8 1.0

Sepal.Length

O OO
1.0

OOOO O
OOOO O O
0.8

o o o
Petal.Width

0.6

O OOOOOO'O
OO O O O
0.4
0.2

0
O OOO O
O OOOOOO O

0.0 0.2 0.4 0.6 0.8 1.0

Petal.Length

12. Plot to see now Petal.Lengtn and Petal.Widfh data Ooints have been distributed originally as per “class” attribule in dataset
 1.0

OOOO O
0.8

OO O OO O O O
Petal.Width

0.6

0 O OOO 0
O OOOOQO0
OO
0.4

O O O OO
0.2

0.0 0.2 0.4 0.6 O.8 1.0

red iltC(lii5tel’ <- at. f a l t o I ( I t ’al?, ñas?tr)

library(qQploL2,

fi harming: prckagt 'g:ploLz' was Lu l under R version 3.3.3

13. Install the package ggplot2 and import it and Plot the cluster results using ggplol
ggplot(irir.new, aes(Petal.fenj-h, Metal.Width, color = esult6cluster}) + gecm pair.K(,

plot(iris.ncw}c("Sepal.Length”, ”Scpal.Wiotn")], col=rosult?cluctcr)

1.00 — • ••

0.75 — •• • •

result8cluster
• 1
0/>0- • •••••••
• 2

0.25 —

0.00 - • ••

0.00 0.25 0.50 075 1.00

Petal.Length
 1.0
0.8

O O

O O O O
SepaI.Width

0.6

D o Do o D O O0
O O O O O O O O O O O
0 0 0 0 0 0
oD
0.4

D O0 0 0 O O O 0 O O O0 O O 0 0
OO O0 0 0 O 0 O
0 0 0 0 0 0 0 0 O
D 0 0 O
O
0.2

0 0 0
0 O O
O O

0.0 0.2 0.4 0.6 0.8 1.0

Sepal.Length
ggplor (irir.rew, ate(Petal.Le.g-h, Petal.Win k, color = result c4ust )) * gecm_poir.t(:

14. Display the clustering results with all parameters

1.00 — • ••

0.75 — •• • •

result8cluster
• 1
0/>0- • •••••••
• 2

0.25 —

0.00 - • ••

0.00 0.25 0.50 075 1.00

Petal.Length
 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0J 0.4 0.6 0.8 1.0

Sepal.Length ° ay, o

TO
0.0 O.2 0,4 0.8 0.8 1.0
Sepal.Width

k
0
’
O
Petal.Width

0
0,0 0.2 0.4 O.6 0.8 1.0

Petal.Length

taLlo(rosu-tSC-ustcr,iris.class) 5Psu2t of tab2P shows that C2uste f

.OispIay.the resorts-in taf›le , ,- ro

## iris.cIas*
fi setosa versicolor virginica
#i 1 0 47 14
#1 2 0 3 36
fi 3 50 0

library(animation)
kJ <-kmezns.zni(iris.new, ?)

Total number of correctly classified instances are: 36 + 47 50= 133 Total number of incorrectly classified instances are: 3 + 14= 17
Accuracy = 133/(133+17) = 0.88 i.e our model has achieved 88% accuracy! In order to Improve this accuracy further, we may try different
values of k

=————————=——————————————————=——————————————————=——————————————————=———————=== K means algoritnms with Animation

16. Disp|ay fhe K Means Algorithm with Animation and visualize the changes in the cluster center
 SepaI.Width SepaI.Width

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

0
.

0.0
0

Sep
SeLpa

L
a

0
06

6
Le

Le
+ng

n
h

+
g
h
o >
4

4
Find cluster? Move centers!
0

0
 SepaI.Width SepaI.Width

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

0
.

0.0
0

Sep
SeLpa

L
a

0 O OØ O
06

6
Le

Le
+ng

n
h

+
g
h
D O O
4

4
Find cluster? Move centers!
0

0
 SepaI.Width SepaI.Width

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

0
.

0.0
0

Sep
SeLpa

L
a

0
06

6
Le

Le
+ng

n
h

+
g
h
D O O
4

4
Find cluster? Move centers!
0

0
 1.0
0.8

Move centers!
SepaI.Width

0.6
0.4

O
0.2

0.0 0.2 0.4 0.6 0.8 1.0

Sepal. Length
1.0
0.8
SepaI.Width

0.6
0.4
0.2

0.0 0.2 0.4 0.6 0.8 1.0

Sepal. Length

17. Impart !actoextra package and visualize the cluster resuIt

library› ':’r rix -'

-- Welcome! Want to learn more° See two - r'-o: x- -related books at ’l

 ’§ sJaĘuao %ńPO ’? T T !1G2Ł\ - g/
'p = szaąuao O<au’siz;)suPawy -> pq
'g = szaąuao %<au’siz;) susan -> pp
' - szaąuao '^au’siz;) suPau -> ą

-0'

Jajsnjn
60t
Dim2

f9
joid ‹ajsnjg
 k=2 k =3

cluster
cluster
M 1
1
2

2 0 2 2 0
Dim1 (739») Dim1 (73%)

k =4 k=5

cluster
cluster
1
1-
2

-2 0 2 -2 0 2
Dim1 (73%) Dim1 (73'f’e)

Conclusion

K means clustering algorithm with ihis dataset is executed and visualized using various r libraries like ggplot,animation,gridExtra,facfoextra.