JRSSEM 2023, Vol. 02 No. 9, 1968 – 1988
E-ISSN: 2807 - 6311, P-ISSN: 2807 - 6494
DOI: 10.59141/jrssem.v2i09.415 https://jrssem.publikasiindonesia.id/index.php/jrssem
NUTRITION CLASSIFICATION IN TODDLERS AT UPTD
PUSKESMAS TIGARAKSA USING A COMPARISON OF
SUPPORT VECTOR MACHINE (SVM) AND K-NEAREST
NEIGHBOR (KNN) METHODS
Amelia Sholikhaq
1
Gerry Firmansyah
2
Bob Tjahjono
3
Habibullah Akbar
4
1,2,3,4
Master of Computer Science Study Program, Faculty of Computer Science
Esa Unggul University Jakarta, Indonesia
*
Email: ameliasholihah[email protected], *gerry@esaunggul.ac.id
budi.tjahjono@esaunggul.ac.id, habibullah.akbar@esaunggul.ac.id
*Correspondence: ameliasholihah8@gmail.com
Submitted
: March 27
th
2023
Revised
: April 12
th
2023
Accepted
: April 20
th
2023
Abstract: Toddlers are a group of people who are vulnerable to nutritional problems. If the
incidence of malnutrition is not addressed, it will hurt children under five, malnutrition is a condition
experienced by a person due to a lack of nutritional intake of the number of nutrients consumed
below. Health centers are required to improve and organize health services as well as possible
therefore researchers conduct research at the UPTD Tigaraksa Health Center by doing a comparison
of classification results on toddler nutritional data using the Support Vector Machine and K-Nearest
Neighbor methods using WEKA Tools. Based on the result of a comparison between the Support
Vector Machine and K-Nearest Neighbor methods using WEKA Tools by carrying out 5 (five) stages
of testing namely: Use Training Set, 4 Cross-Validation, 8 Cross-Validation, 50% Percentage Split
dan 80% Percentage Split, the results show that the Support Vector Machine method Kernel Radial
Basis Function (RBF) is an average accuracy value of 100% higher than the K-Nearest Neighbor
Euclidean Distance algorithm with an average accuracy of 93%.
Keywords: Toddler Nutrition; Data Mining; Classification; Support Vector Machine and K-Nearest
Neighbor methods; WEKA Tools.
Amelia Sholikhaq
1
Gerry Firmansyah
2
Budi Tjahjono
3
Habibullah Akbar
4
| 1969
INTRODUCTION
Poor nutrition is a condition
experienced by a person due to a lack of
nutritional intake or the number of
nutrients consumed below standard.
Nutrients needed include carbohydrates,
proteins, and calories. One of the most
important and common nutritional
problems experienced by infants under 5
years old (toddlers) is a lack of protein
energy. It is associated with the economic
level of society. In addition, parents lack
knowledge about the importance of
nutrition for children's growth and
development. The nutritional status of
toddlers can be determined through
laboratory examination or anthropometry.
Anthropometric measurements are
measurements used to determine a
person's nutritional state.
Puskesmas is a technical
implementation unit of the district/city
health office that is responsible for
organizing health development in a work
area. Puskesmas are required to improve
and provide health services as well as
possible. UPTD Puskesmas Tigarakasa is a
Puskesmas located in Tigaraksa District,
Tangerang Regency, which is located at Jl.
Kongsi No.12, Tigaraksa Village, Tigaraksa
District, Tangerang Regency, Banten
Province, Puskesmas work area. Tigaraksa
Health Center covers several villages,
namely Bantar Panjang, Cileles, Kadu
Agung, Margasari, Sodong, Tapos, and
Tigaraksa.
In this study, the goal to be achieved is
to improve the accuracy of results using
method comparison with the comparison
of Support Vector Machine and K-Nearest
Neighbors methods in classifying nutrition
in toddlers (Arsi & Waluyo, 2021). Proving
that using a comparison of the Support
Vector Machine and K-Nearest Neighbors
methods can increase the percentage of
accuracy so that it is more optimal in the
early detection of nutritional status in
toddlers quickly, precisely, and accurately
(Sugara & Subekti, 2019).
MATERIALS AND METHODS
Nutritional Status of Toddlers
Nutritional Status is a measure of
success in fulfilling nutrition for children
indicated by the child's weight and height.
Nutritional status can be defined as the
health status produced by the balance
between nutritional needs and inputs. A
child can be said to be a toddler when he is
between 0 to 5 years old. Because the term
BALITA is an abbreviation of Infants Under
Five Years". The age of toddlers is the age of
growth, which is when a toddler must be
active and energetic in acting. Toddlers are
active and energetic in doing actions
because curiosity about something they
meet arises in their minds. The nutritional
status of toddlers can be classified as
follows: Poor Nutrition, Lack of Nutrition,
Good Nutrition, and More Nutrition (Hariri
&; Pamungkas, 2016).
Data Mining
Data mining is the process by which
statistical, mathematical, artificial
intelligence and machine learning
techniques are used to extract and identify
useful information and related knowledge
in large databases. Data mining is not an
entirely new field. One of the difficulties in
1970 | Nutrition Classification In Toddlers at UPTD Puskesmas Tigaraksa Using A Comparison of
Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) Methods
identifying data mining is the fact that data
mining has long roots in fields such as
artificial intelligence, machine learning,
statistics, databases, and information
retrieval (Nikmatun & Waspada, 2019).
Tools WEKA
WEKA has useful tools for data
preprocessing, classification, regression,
clustering, association rules, and
visualization. Can be used to preprocess
data, enter into a learning schema, and
analyze the classification generated by its
performance, done without writing
program code. Examples of using WEKA by
applying a learning method to a dataset
and analyzing the results to obtain
information about the data, or applying the
method and comparing its performance to
be selected.
Classification
Classification is a data mining method
that can be used for the process of
searching for a set of data models
(functions) that can explain and distinguish
data classes or data concepts, which has the
aim that the model can be used to predict
class objects that have labels whose value is
unknown or used to predict the tendency of
data that often appears in the future.
The model in classification is defined in
detail as a working model whose process of
conducting training / requires a learning
model of the target function, which is
usually interpreted as a place to receive
input (training data), then be able to think
about the input, and provide answers as an
output of the results of his thoughts. The
model is used to predict classes from test
data. The process of work in classification
can be seen in Figure 1.
Figure 1. The process of classification work
In Figure 1. The model that is already
built at the time of training input can then
be used to predict class labels from new
data that does not yet know the class labels.
In building a model during the training
process, an algorithm is needed to build it,
which is called a training algorithm
(learning algorithm).
Comparisons
In the Big Indonesian Dictionary (KBBI)
Comparison is the difference (difference)
between similarities and similarities, then
the comparison is an attempt to observe
the differences or similarities possessed by
two or more objects that have a certain
similarity (Nasution & Hayaty, 2019;
Pratama & Salamah, 2022).
Metode Support Vector Machine (SVM)
Support Vector Machine (SVM) is one
of the existing classification methods
(Fadilah et al., 2020). SVM was developed
by Boser, Guyon, and Vapnik, and was first
presented in 1992 at the Annual Workshop
on Computational Learning Theory
(Permana & Sahara, 2019). The basic
concept of SVM is a harmonized
combination of computational theories of
money that had existed decades earlier,
such as the hyperplane margin (Duda and
Hart in 1973, cover in 1965, Vapnik in 1964,
Amelia Sholikhaq
1
Gerry Firmansyah
2
Budi Tjahjono
3
Habibullah Akbar
4
| 1971
etc.), the kernel introduced by Aronszajn in
1950, and so with other supporting
concepts (Wati et al., 2020). However, until
1992, there had never been an attempt to
assemble these components (Mase et al.,
2018).
Metode K-Nearest Neighbors (KNN)
The K-Nearest Neighbors (KNN)
method is a method for classifying based
on the proximity of the location (distance)
of one data with other data (Pamungkas &
Kharisudin, 2021). K-Nearest Neighbors
includes supervised learning algorithms
(Saidah et al., 2019). The working principle
of K-Nearest Neighbors (KNN) itself is to
find the closest distance between the
evaluated data and (K) its closest neighbor
in the training data. Before finding the
closest distance between the evaluated
data, the K-Nearest Neighbors algorithm
must be preprocessed or normalized first
(Budianto et al., 2019).
Testing and Evaluation
The test of the analysis aims to
determine the level of accuracy of the
comparison between the support vector
machine method and the k-nearest
neighbor in the classifier of determining
nutritional status in toddlers at UPTD
Puskesmas Tigaraksa (Baita et al., 2021).
Accuracy testing with classification is
carried out in several experiments to get
more accurate results (Hakim et al., 2020).
In processing the classification,
researchers use the Weka Tool to determine
a better level of accuracy between the
Support Vector Machine method and K-
Nearest Neighbor with several stages in
processing datasets in each classification
method using the Support Vector Machine
and K-Nearest Neighbor (Ichwan & Dewi,
2018), namely:
a. Use Training Set, at this stage WEKA
Tools uses the previously inputted training
data as testing data. In other words, the
training and testing process uses the same
data.
b. Cross-Validation, the training data will
randomly be divided into k parts at this
stage. Furthermore, the k-1 part is used as
training data, and one part is used as test
data. The process is repeated so that each
part has the opportunity to become test
data. In WEKA Tools, the default value is 10.
c. Percentage Split, the data inputted in
the previous step will be divided into
training data and test data based on a
certain percentage. The default value in
WEKA Tools is 66 %, where the input data is
divided into 66% training data and 34% test
data.
Data Collection
In collecting toddler data used in this
study, data obtained from UPTD Puskesmas
Tigaraksa data on toddlers born in 2018 to
2022 aged 0 months to 60 months,
focusing on variables or features that are
numerical in nature with four parameters,
namely Age in months, Gender, Weight in
kg, Height in cm.
1972 | Nutrition Classification In Toddlers at UPTD Puskesmas Tigaraksa Using A Comparison of
Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) Methods
Table 1. Data on Toddlers UPTD Puskesmas Tigaraksa toddler age aged 5 to 59 months
RESULTS AND DISCUSSION
Nutritional Status Data for Toddlers Using Z
Score
In inputting anthropometric data for
toddlers aged between 1 to 59 months
carried out by UPDT Puskesmas Tigaraksa
which includes seven villages, namely
tigaraksa, kaduagung, marga sari, song,
tapos, bantarpanjang, and cells, which have
been calculated using the z-score formula
to determine the nutritional status of each
toddler who has been calculated according
to age, gender, weight (kg) and
length/height (cm).
ID NAMA JK USIA-BLN ATT-BB/U ATT-TB/U ATT-TB/BB
1 AWAL Laki-laki 5
-0,07 -1,30 -0,04
2 ROBI Laki-laki 8
-1,13 -1,85 -0,24
3 ALIY Laki-laki 7
-2,24 -1,91 -2,94
4 DEFR Laki-laki 10
-2,20 -1,52 -2,08
5 ERIN Laki-laki 5
1,50 -2,00 1,01
6 GEND Laki-laki 11
2,60 -1,30 1,07
7 ARAM Laki-laki 6 2,60
-1,06 2,12
8 LUTH Laki-laki 8 2,66
-0,54 2,18
9 MUHA Laki-laki 10 2,79
-1,17 3,97
10 RIFAT Laki-laki 9 3,20
-1,63 3,17
… ….. …. …. ….
…. ….
… ….. …. …. ….
…. ….
… ….. …. …. ….
…. ….
91 ALFA Perempuan 56 -1,69 -1,61 -1,1
92 RINA Perempuan 55 -1,85 -1,56 -1,3
93 ADIB Perempuan 52
-3,77 -2,32 -2,4
94 AFIF Perempuan 51
-2,46 -2,13 -2,3
95 ARSY Perempuan 50 1,24
0,86 1,5
96 ASMA Perempuan 53 1,17 2,56 1,5
97 AURE Perempuan 58 2,45
-0,65 2,2
98 AYUN Perempuan 50 2,55
-0,87 2,2
99 CHAT Perempuan 58 3,34
-1,22 3,4
100 FITR Perempuan 52 3,22
-1,11 3,2
Amelia Sholikhaq
1
Gerry Firmansyah
2
Budi Tjahjono
3
Habibullah Akbar
4
| 1973
Table 2. UPTD Toddler Data Puskesmas Tigaraksa
In the table above the toddler data of
UPTD Puskesmas Tigaraksa, researchers
took samples of 100 toddler data for the
August 2022 period aged 1 to 59 months.
Among them are males and females, whose
grouping can be seen in the table below.
Table 3. Toddler data grouped by gender
No
Gender
Sum
Status
1
Man
10
Undernutrition
10
Good Nutrition
10
More Nutritional
Risks
10
More Nutrition
10
Obesity
2
Woman
10
Undernutrition
10
Good Nutrition
ID NAMA ATT-JK USIA-BLN ATT-BB/U BB/U ATT-TB/U TB/U ATT-TB/BB STATUS
1 AWAL Laki-laki 5
-0,07 BB Normal -1,30 TB Normal -0,04 Gizi Baik
2 ROBI Laki-laki 8
-1,13 BB Normal -1,85 TB Normal -0,24 Gizi Baik
3 ALIY Laki-laki 7
-2,24 BB Kurang -1,91 TB Normal -2,94 Gizi Kurang
4 DEFR Laki-laki 10
-2,20 BB Kurang -1,52 TB Normal -2,08 Gizi Kurang
5 ERIN Laki-laki 5
1,50 BB Resiko Lebih -2,00 TB Normal 1,01 Resiko Gizi Lebih
6 GEND Laki-laki 11
2,60 BB Resiko Lebih -1,30 TB Normal 1,07 Resiko Gizi Lebih
7 ARAM Laki-laki 6 2,60 BB Lebih
-1,06 TB Normal 2,12 Gizi Lebih
8 LUTH Laki-laki 8 2,66 BB Lebih
-0,54 TB Normal 2,18 Gizi Lebih
9 MUHA Laki-laki 10 2,79 BB Obesitas
-1,17 TB Normal 3,97 Obesitas
10 RIFAT Laki-laki 9 3,20 BB Obesitas
-1,63 TB Normal 3,17 Obesitas
11 ELSA Perempuan 5 -0,19 BB Normal
-1,06 TB Normal 0,35 Gizi Baik
12 LALA Perempuan 8 0,08 BB Normal
-1,56 TB Normal 0,71 Gizi Baik
13 FELI Perempuan 7 -2,76 BB Kurang
-1,77 TB Normal -2,20 Gizi Kurang
14 DITA Perempuan 10 -2,85 BB Kurang
-1,89 TB Normal -2,11 Gizi Kurang
15 AKIL Perempuan 5
2,20 BB Resiko Lebih -1,52 TB Normal 1,08 Resiko Gizi Lebih
16 HAND Perempuan 11
2,48 BB Resiko Lebih -1,20 TB Normal 1,10 Resiko Gizi Lebih
17 ANIA Perempuan 6 2,60 BB Lebih
-0,39 TB Normal 2,18 Gizi Lebih
18 ROTU Perempuan 8 2,66 BB Lebih
-0,84 TB Normal 2,19 Gizi Lebih
19 ZIHA Perempuan 11 3,20 BB Obesitas
-1,88 TB Normal 3,67 Obesitas
20 RANI Perempuan 9 3,26 BB Obesitas
-1,85 TB Normal 3,19 Obesitas
… …. …. …. …. …..
…. ….. …. …..
… …. …. …. …. …..
…. ….. …. …..
… …. …. …. …. …..
…. ….. …. …..
81 BILA Laki-laki 58 -1,04 BB Normal 2,57 TB Tinggi 0,1 Gizi Baik
82 ALFI Laki-laki 50 -1,33 BB Normal 2,20 TB Tinggi -0,5 Gizi Baik
83 CHAR Laki-laki 53
-2,68 BB Kurang -2,32 TB Pendek -2,5 Gizi Kurang
84 ALTA Laki-laki 50
-2,55 BB Kurang -2,26 TB Pendek -2,4 Gizi Kurang
85 AZZA Laki-laki 49 1,18 BB Resiko Lebih 2,22 TB Tinggi 1,5 Resiko Gizi Lebih
86 EDWA Laki-laki 56 1,21 BB Resiko Lebih
-0,48 TB Normal 1,5 Resiko Gizi Lebih
87 EVAN Laki-laki 58 2,22 BB Lebih
-1,27 TB Normal 2,9 Gizi Lebih
88 FARZA Laki-laki 59 2,80 BB Lebih
-1,35 TB Normal 2,5 Gizi Lebih
89 GIBR Laki-laki 49 3,45 BB Obesitas
-1,09 TB Normal 3,3 Obesitas
90 LINT Laki-laki 55 3,33 BB Obesitas
-0,06 TB Normal 4,2 Obesitas
91 ALFA Perempuan 56 -1,69 BB Normal -1,61 TB Normal -1,1 Gizi Baik
92 RINA Perempuan 55 -1,85 BB Normal -1,56 TB Normal -1,3 Gizi Baik
93 ADIB Perempuan 52
-3,77 BB Kurang -2,32 TB Pendek -2,4 Gizi Kurang
94 AFIF Perempuan 51
-2,46 BB Kurang -2,13 TB Pendek -2,3 Gizi Kurang
95 ARSY Perempuan 50 1,24 BB Resiko Lebih
0,86 TB Normal 1,5 Resiko Gizi Lebih
96 ASMA Perempuan 53 1,17 BB Resiko Lebih 2,56 TB Tinggi 1,5 Resiko Gizi Lebih
97 AURE Perempuan 58 2,45 BB Lebih
-0,65 TB Normal 2,2 Gizi Lebih
98 AYUN Perempuan 50 2,55 BB Lebih
-0,87 TB Normal 2,2 Gizi Lebih
99 CHAT Perempuan 58 3,34 BB Obesitas
-1,22 TB Normal 3,4 Obesitas
100 FITR Perempuan 52 3,22 BB Obesitas
-1,11 TB Normal 3,2 Obesitas
1974 | Nutrition Classification In Toddlers at UPTD Puskesmas Tigaraksa Using A Comparison of
Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) Methods
10
More Nutritional
Risks
10
More Nutrition
10
Obesity
Table 4. Data on the nutritional status of toddlers before normalization are grouped
according to age
In the table above, toddler nutrition
data has been normalized with the
following information:
1. Gender (male = 1, female = 2).
2. Age is calculated in months according to
the main data of the toddler table.
3. The Weight Score uses the z-score value
of body weight against age (BB/U).
ID ATT-JK USIA-BLN ATT-BB/U ATT-TB/U STATUS
1 Laki-laki 5
-0,07 -1,30 Gizi Baik
2 Laki-laki 8
-1,13 -1,85 Gizi Baik
3 Laki-laki 7
-2,24 -1,91 Gizi Kurang
4 Laki-laki 10
-2,20 -1,52 Gizi Kurang
5 Laki-laki 5
1,50 -2,00 Resiko Gizi Lebih
6 Laki-laki 11
2,60 -1,30 Resiko Gizi Lebih
7 Laki-laki 6 2,60
-1,06 Gizi Lebih
8 Laki-laki 8 2,66
-0,54 Gizi Lebih
9 Laki-laki 10 2,79
-1,17 Obesitas
10 Laki-laki 9 3,20
-1,63 Obesitas
… …… …… …… …… ……
… …… …… …… …… ……
… …… …… …… …… ……
81 Laki-laki 58 -1,04 2,57 Gizi Baik
82 Laki-laki 50 -1,33 2,20 Gizi Baik
83 Laki-laki 53
-2,68 -2,32 Gizi Kurang
84 Laki-laki 50
-2,55 -2,26 Gizi Kurang
85 Laki-laki 49 1,18 2,22 Resiko Gizi Lebih
86 Laki-laki 56 1,21
-0,48 Resiko Gizi Lebih
87 Laki-laki 58 2,22
-1,27 Gizi Lebih
88 Laki-laki 59 2,80
-1,35 Gizi Lebih
89 Laki-laki 49 3,45
-1,09 Obesitas
90 Laki-laki 55 3,33
-0,06 Obesitas
91 Perempuan 56 -1,69 -1,61 Gizi Baik
92 Perempuan 55 -1,85 -1,56 Gizi Baik
93 Perempuan 52
-3,77 -2,32 Gizi Kurang
94 Perempuan 51
-2,46 -2,13 Gizi Kurang
95 Perempuan 50 1,24
0,86 Resiko Gizi Lebih
96 Perempuan 53 1,17 2,56 Resiko Gizi Lebih
97 Perempuan 58 2,45
-0,65 Gizi Lebih
98 Perempuan 50 2,55
-0,87 Gizi Lebih
99 Perempuan 58 3,34
-1,22 Obesitas
100 Perempuan 52 3,22
-1,11 Obesitas
Amelia Sholikhaq
1
Gerry Firmansyah
2
Budi Tjahjono
3
Habibullah Akbar
4
| 1975
4. The Height Value still uses the z-score
value of height against age (TB/U).
Nutritional status results from the
calculation of z-score body weight to
height (BB / TB).
Classification of Toddler Nutrition Data
Using Support Vector Machine (SVM)
Method and K-Nearest Neighbor Using
Weka Tool
a. Using the Support Vector Machine
(SVM) method.
Classification of the nutritional status
of toddlers using the support vector
machine (SVM) method with Radial Basis
Function (RBF) kernels is carried out with 5
tests, namely:
Use Training Set (data testing with the
same training data)
4 &; 8 Cross-Validation (dividing data
into k-subsets. For example Folds are
used 10, 9 will be used as training data,
and 1 as testing data until all data)
50% Percentage Split. (Splits the data
according to the parameters that will
be the data training).
80% Percentage Split. (Splits the data
according to the parameters that will
be the data training).
The following is the result of the
support vector machine (SVM)
classification using the WEKA tool:
1. Results of Support Vector Machine
Classification Using WEKA Tool (Use
Data Training)
Gambar 2. Klasifikasi Support Vector Machine (SVM) use data training (Naufal et al., 2020)
The picture above is the result of the
classification of support vector machines in
the WEKA tool using a use training set
which shows the results of 100 correct
predictions with an accuracy of 100% and 0
incorrect predictions with a percentage of
0% with a classification time of 0.02
seconds (Muhammad Yusuf Ramadan,
2019).
Table 5. Confusion Matrix on WEKA Use Data Training Tools
Predictions
Good
Nutrition
Undernutrition
More Risk
More
Nutrition
Obesity
1976 | Nutrition Classification In Toddlers at UPTD Puskesmas Tigaraksa Using A Comparison of
Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) Methods
Good Nutrition
20
0
0
0
0
Undernutrition
0
20
0
0
0
More
Nutritional
Risks
0
0
20
0
0
More Nutrition
0
0
0
20
0
Obesity
0
0
0
0
20
Based on the table above, it can be
seen that the nutritional status of toddlers
can be classified correctly, namely 100 data
on the nutritional status of toddlers.
Therefore, an accuracy value of 100% is
obtained.
2. Results of Support Vector Machine
Classification Using WEKA Tool (4 Cross-
Validation)
Figure 3. Support Vector Machine Classification (4 Cross-Validation)
The picture above is the result of the
classification of the support vector machine
in the WEKA tool using 4 Cross-Validation
which shows the results of 100 correct
predictions with an accuracy of 100% and 0
incorrect predictions with a percentage of
0% with a classification time of 0.07
seconds.
Table 6. Confusion Matrix on 4 Cross-Validation Tools WEKA
Predictions
Good
Nutrition
Undernutrition
More Risk
More
Nutrition
Obesity
Good Nutrition
20
0
0
0
0
Undernutrition
0
20
0
0
0
More
Nutritional
Risks
0
0
20
0
0
More Nutrition
0
0
0
20
0
Obesity
0
0
0
0
20
Amelia Sholikhaq
1
Gerry Firmansyah
2
Budi Tjahjono
3
Habibullah Akbar
4
| 1977
DOI: 10.59141/jrssem.v2i09.415 https://jrssem.publikasiindonesia.id/index.php/jrssem
Based on the table above, it can be
seen that the nutritional status of toddlers
can be calcified correctly, namely 100 data
from 100 data on the nutritional status of
toddlers. Therefore, an accuracy value of
100% is obtained.
3. Results of Support Vector Machine
Classification Using WEKA Tool (8 Cross-
Validation)
Figure 4. Support Vector Machine Classification (8 Cross-Validation)
The picture above is the result of the
support vector machine classification in the
WEKA tool using 8 Cross-Validation which
shows the results of 100 correct predictions
with an accuracy of 100% and 0 incorrect
predictions with a percentage of 0% with a
classification time of 0.04 seconds.
Table 7. Confusion Matrix on 8 WEKA Cross-Validation Tools
Predictions
Good
Nutrition
Undernutrition
More Risk
More
Nutrition
Obesity
Good Nutrition
20
0
0
0
0
Undernutrition
0
20
0
0
0
More
Nutritional
Risks
0
0
20
0
0
More Nutrition
0
0
0
20
0
Obesity
0
0
0
0
20
Based on the table above, it can be
seen that the nutritional status of toddlers
can be calcified correctly, namely 100 data
from 100 data on the nutritional status of
toddlers. While none of the data is
classified less precisely or differently from
the original data. Therefore, an accuracy
value of 100% is obtained.
4. Support Vector Machine classification
results using WEKA Tool (50%
Percentage Split).
Amelia Sholikhaq
1
Gerry Firmansyah
2
Budi Tjahjono
3
Habibullah Akbar
4
| 1978
DOI: 10.59141/jrssem.v2i09.415 https://jrssem.publikasiindonesia.id/index.php/jrssem
Figure 5. Support Vector Machine Classification (50% Percentage Split)
The picture above is the result of the
support vector machine classification in the
WEKA tool using a 50% Percentage Split
which shows the results of 50 correct
predictions with 100% accuracy and 0
incorrect predictions with a percentage of
0% with a classification time of 0.02
seconds.
Table 8. Confusion Matrix at 50% Percentage Split with WEKA Tools
Predictions
Good
Nutrition
Undernutrition
More Risk
More
Nutrition
Obesity
Good Nutrition
11
0
0
0
0
Undernutrition
0
14
0
0
0
More
Nutritional
Risks
0
0
11
0
0
More Nutrition
0
0
0
7
0
Obesity
0
0
0
0
7
Based on the table above, it can be
seen that the nutritional status of toddlers
can be classified correctly, namely 50 data
from 50 nutritional status data of toddlers.
While none of the other data is classified
less precisely or differently from the
original data. Therefore, an accuracy value
of 100% is obtained.
5. Support Vector Machine classification
results using WEKA Tool (80%
Percentage Split).
Amelia Sholikhaq
1
Gerry Firmansyah
2
Budi Tjahjono
3
Habibullah Akbar
4
| 1979
DOI: 10.59141/jrssem.v2i09.415 https://jrssem.publikasiindonesia.id/index.php/jrssem
Figure 6. Support Vector Machine Classification (80% Percentage Split)
The picture above is the result of the
support vector machine classification in the
WEKA tool using an 80% Percentage Split
which shows the results of 20 correct
predictions with 100% accuracy and no
wrong predictions with a percentage of 0%
with a classification time of 0.03 seconds.
Table 9. Confusion Matrix at 80% Percentage Split with WEKA Tools
Predictions
Good
Nutrition
Undernutrition
More Risk
More
Nutrition
Obesity
Good Nutrition
3
0
0
0
0
Undernutrition
0
7
0
0
0
More
Nutritional
Risks
0
0
4
0
0
More Nutrition
0
0
0
2
0
Obesity
0
0
0
0
4
Based on the table above, it can be
seen that the nutritional status of toddlers
can be classified correctly, namely 20 data
from 20 nutritional status data of toddlers.
While no other data is classified less
precisely or different from the original data.
Therefore, an accuracy value of 100% is
obtained.
b. Comparison Results of Support Vector
Machine Accuracy Evaluation from
Nutritional Status Data in Toddlers
After analyzing the Support Vector
Machine classification in the WEKA tool
using the Use Training Set, 4 Fold Cross
Validation, 8 Fold Cross Validation, 50%
Percentage Split, and 80% Percentage Split,
the accuracy obtained in each test has the
same value, namely with an accuracy
percentage of 100% for Correctly Classified
Instances and 0% for Incorrectly Classified
Instances. The comparison can be seen in
Table 10.
Amelia Sholikhaq
1
Gerry Firmansyah
2
Budi Tjahjono
3
Habibullah Akbar
4
| 1980
DOI: 10.59141/jrssem.v2i09.415 https://jrssem.publikasiindonesia.id/index.php/jrssem
Table 10. Support Vector Machine Accuracy Evaluation Comparison
Evaluation Model
Accuracy
Number
of
Toddlers
Percentage
Use Training Set
Correctly Classified Instances
100
100%
Incorreclty Classified
Instances
0
0%
4 Fold Cross-Validation
Correctly Classified Instances
100
100%
Incorreclty Classified
Instances
0
0%
8 Fold Cross-Validation
Correctly Classified Instances
100
100%
Incorreclty Classified
Instances
0
0%
50% Percentage Split
Correctly Classified Instances
50
100%
Incorreclty Classified
Instances
0
0%
70% Percentage Split
Correctly Classified Instances
20
100%
Incorreclty Classified
Instances
0
0%
c. Menggunakan Metode K-Nearest
Neighbour (KNN).
Classification of nutritional status of
toddlers using K-Nearest Neighbor
with Euclidean Distance algorithm
carried out 5 tests (Nikmatun &
Waspada, 2019), namely:
Use Training Set (data testing with the
same training data)
4 &; 8 Cross-Validation (dividing data
into k-subsets. For example Folds used
10.9 will be used as training data and 1
as testing data up to all data)
50% &; 80% Percentage Split. (Splits
the data according to the parameters
that will be the data training).
The following results from the K-
Nearest Neighbor classification using the
WEKA (Use Data Training) Tool.
1. Hasil Klasifikasi K-Nearest Neighbour
Menggunakan Tool WEKA (Use Data
Training).
Amelia Sholikhaq
1
Gerry Firmansyah
2
Budi Tjahjono
3
Habibullah Akbar
4
| 1981
DOI: 10.59141/jrssem.v2i09.415 https://jrssem.publikasiindonesia.id/index.php/jrssem
Gambar 7. Klasifikasi K-Nearest Neighbour (Use Training Set)
The picture above is the result of the
K-Nearest Neighbor classification in the
WEKA tool using a use training set which
results in 100 correct predictions with 100%
accuracy and no wrong predictions with a
percentage of 0% with a classification time
of 0.03 seconds (Amalia et al., 2021).
Table 11. Confusion Matrix on Using Data Training with WEKA Tools
Predictions
Good
Nutrition
Undernutrition
More Risk
More
Nutrition
Obesity
Good Nutrition
20
0
0
0
0
Undernutrition
0
20
0
0
0
More Nutritional
Risks
0
0
20
0
0
More Nutrition
0
0
0
20
0
Obesity
0
0
0
0
20
Based on the table above, it can be
seen that the nutritional status of toddlers
can be classified correctly, namely 100 data
from 100 nutritional status data of toddlers.
While no other data is classified less
precisely or different from the original data.
Therefore, an accuracy value of 100% is
obtained.
2. K-Nearest Neighbor Classification
Results Using WEKA Tool (4 Cross-
Validation)
Amelia Sholikhaq
1
Gerry Firmansyah
2
Budi Tjahjono
3
Habibullah Akbar
4
| 1982
DOI: 10.59141/jrssem.v2i09.415 https://jrssem.publikasiindonesia.id/index.php/jrssem
Gambar 8. Klasifikasi K-Nearest Neighbour (4 Cross-Validation)
The picture above is the result of the K-
Nearest Neighbor classification in the WEKA
tool using 4 Cross-Validation which shows
the results of 89 correct predictions with an
accuracy of 89% and 11 incorrect
predictions with a percentage of 11% with
a classification time of 0 seconds.
Table 12. Confusion Matrix in 4 Cross-Validation with WEKA Tools
Predictions
Good
Nutrition
Undernutrition
More Risk
More
Nutrition
Obesity
Good Nutrition
18
0
0
0
0
Undernutrition
0
18
2
0
0
More Nutritional
Risks
0
1
18
1
0
More Nutrition
0
0
3
17
0
Obesity
0
0
2
0
18
Based on the table above, it can be
seen that the nutritional status of toddlers
can be calcified correctly, namely 89 data
from 100 nutritional status data of toddlers.
While the other 11 data are classified
incorrectly or differently from the original
data. Therefore, an accuracy value of 89%
was obtained.
3. The results of the K-nearest neighbor
classification using the WEKA (8 Cross-
Validation) tool.
Amelia Sholikhaq
1
Gerry Firmansyah
2
Budi Tjahjono
3
Habibullah Akbar
4
| 1983
DOI: 10.59141/jrssem.v2i09.415 https://jrssem.publikasiindonesia.id/index.php/jrssem
Gambar 9. Klasifikasi K-Nearest Neighbour (8 Cross-Validation)
The picture above is the result of the K-
Nearest Neighbor classification in the WEKA
tool using 8 cross-validations which shows
the results of 90 correct predictions with an
accuracy of 90% and 10 incorrect
predictions with a percentage of 10% with
a classification time of 0 seconds.
Table 13. Confusion Matrix in 8 Cross-Validation with WEKA Tools
Predictions
Good
Nutrition
Undernutrition
More Risk
More
Nutrition
Obesity
Good Nutrition
18
0
0
0
2
Undernutrition
0
18
0
0
0
More Nutritional
Risks
0
1
18
1
0
More Nutrition
0
0
2
18
0
Obesity
0
0
2
0
18
Based on the table above, it can be
seen that the nutritional status of toddlers
can be classified correctly, namely 90 data
from 100 nutritional status data of toddlers.
While the other 10 data are classified
incorrectly or differently from the original
data. Therefore, an accuracy value of 90% is
obtained.
4. K-Nearest Neighbor Classification
Results Using WEKA Tool (50%
Percentage Split)
Amelia Sholikhaq
1
Gerry Firmansyah
2
Budi Tjahjono
3
Habibullah Akbar
4
| 1984
DOI: 10.59141/jrssem.v2i09.415 https://jrssem.publikasiindonesia.id/index.php/jrssem
Gambar 10. Klasifikasi K-Nearest Neighbour (50% Percentage Split)
The picture above is the result of the K-
Nearest Neighbor classification in the WEKA
tool using a 50% Percentage Split which
shows the results of 45 correct predictions
with an accuracy of 90% and 5 incorrect
predictions with a percentage of 10% with
a classification time of 0.01 seconds.
Table 14. Confusion Matrix at 50% Percentage Split with WEKA Tools
Predictions
Good
Nutrition
Undernutrition
More Risk
More
Nutrition
Obesity
Good Nutrition
8
1
0
0
2
Undernutrition
1
13
0
0
0
More Nutritional
Risks
0
0
10
1
0
More Nutrition
0
0
0
7
0
Obesity
0
0
0
0
7
Based on the table above, it can be
seen that the nutritional status of toddlers
can be calcified correctly, namely 45 data
from 50 nutritional status data of toddlers.
While the other 5 data are classified
incorrectly or differently from the original
data. Therefore, an accuracy value of 90% is
obtained.
5. K-Nearest Neighbor Classification
Results Using WEKA Tool (80%
Percentage Split)
Amelia Sholikhaq
1
Gerry Firmansyah
2
Budi Tjahjono
3
Habibullah Akbar
4
| 1985
DOI: 10.59141/jrssem.v2i09.415 https://jrssem.publikasiindonesia.id/index.php/jrssem
Gambar 11. Klasifikasi K-Nearest Neighbour (80% Percentage Split)
The picture above is the result of the K-
Nearest Neighbor classification in the WEKA
tool using an 80% Percentage Split which
shows the results of 19 correct predictions
with an accuracy of 95% and 1 wrong
prediction with a percentage of 5% with a
classification time of 0.03 seconds.
Table 15. Confusion Matrix at 80% Percentage Split with WEKA Tools
Predictions
Good
Nutrition
Undernutrition
More Risk
More
Nutrition
Obesity
Good Nutrition
2
0
0
0
1
Undernutrition
0
7
0
0
0
More Nutritional
Risks
0
0
4
0
0
More Nutrition
0
0
0
2
0
Obesity
0
0
0
0
4
Based on the table above, it can be
seen that the nutritional status of toddlers
can be classified correctly, namely 19 data
from 20 nutritional status data of toddlers.
While the other 1 data is classified
incorrectly or differently from the original
data. Therefore, an accuracy value of 95% is
obtained.
d. Comparison of K-nearest neighbor
accuracy evaluation results from
nutritional status data in toddlers
After analyzing the classification of
the K-Nearest Neighbor with the Euclidean
Distance algorithm in the WEKA tool using
the Use Training Set, 4 Fold Cross Validation,
8 Fold Cross Validation, 50% Percentage
Split, and 80% Percentage Split, the highest
accuracy was obtained using the Use
Training Set with an accuracy percentage of
100% for Correctly Classified Instances and
0% for Incorrectly Classified Instances. The
comparison can be seen in Table 16.
1986 | Nutrition Classification In Toddlers at UPTD Puskesmas Tigaraksa Using A Comparison of
Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) Methods
Table 16. K-Nearest Neighbor Accuracy Evaluation Comparison
Evaluation Model
Accuracy
Number
of
Toddlers
Percentage
Use Training Set
Correctly Classified
Instances
100
100%
Incorreclty Classified
Instances
0
0%
4 Fold Cross-
Validation
Correctly Classified
Instances
89
89%
Incorreclty Classified
Instances
11
11%
8 Fold Cross-
Validation
Correctly Classified
Instances
90
90%
Incorreclty Classified
Instances
10
10%
50% Percentage Split
Correctly Classified
Instances
45
90%
Incorreclty Classified
Instances
5
10%
70% Percentage Split
Correctly Classified
Instances
19
95%
Incorreclty Classified
Instances
1
5%
e. Comparison of Support Vector
Machine Method Classification Results
with K-Nearest Neighbor
Comparison of classification results
using 2 methods, namely Support Vector
Machine and K-Nearest Neighbor can be
seen in the table below (Faruk & Nafi’iyah,
2020) (Iskandar & Nataliani, 2021).
Table 17. Comparison of SVM and KNN Classification Results
CONCLUSIONS
Based on the results of research that
we have conducted using data mining,
namely the classification of nutrition in
toddlers at UPTD Puskesmas Tigaraksa
5 tes
SVM
KNN
Test Results
Correctly Classified
Instances
Correctly Classified
Instances
Use Training Set
100%
100%
4 Fold Cross-Validation
100%
89%
8 Fold Cross-Validation
100%
90%
50% Percentage Split
100%
90%
80% Percentage Split
100%
95%
Percentage Average
100%
93%
Amelia Sholikhaq
1
Gerry Firmansyah
2
Budi Tjahjono
3
Habibullah Akbar
4
| 1987
using a comparison of the support vector
machine (SVM) and k-nearest neighbor
(KNN) methods, a conclusion can be drawn,
as follows:
a. Classification using the Support Vector
Machine (SVM) method with the Radial
Basis Function (RBF) kernel doing 5
tests can be obtained the accuracy
value of the five tests has the same
accuracy, which is 100%, where as
many as 100 toddler data are declared
correct (Correctly Classified Instances).
b. Classification using the K-Nearest
Neighbor (KNN) method with the
Euclidean Distance algorithm by
conducting 5 tests using only training
set data obtained an accuracy value of
100%, where as many as 100 toddler
data were declared correct (Correctly
Classified Instances) but the next four
tests could not produce perfect
accuracy scores.
The highest accuracy value in
classifying toddler nutrition data using
comparisons can be seen in that the
Support Vector Machine method provides a
more accurate accuracy value with the
same percentage, which is 100% on each
test tested.
REFERENCES
Amalia, B. S., Umaidah, Y., & Mayasari, R.
(2021). Analisis Sentimen Review
Pelanggan Restoran Menggunakan
Algoritma Support Vector Machine
Dan K-Nearest Neighbor.
SITEKIN:
Jurnal Sains, Teknologi Dan Industri
,
19
(1), 28–34.
Arsi, P., & Waluyo, R. (2021). Analisis
Sentimen Wacana Pemindahan Ibu
Kota Indonesia Menggunakan
Algoritma Support Vector Machine
(SVM).
Jurnal Teknologi Informasi Dan
Ilmu Komputer
,
8
(1), 147.
Baita, A., Pristyanto, Y., & Cahyono, N.
(2021). Analisis Sentimen Mengenai
Vaksin Sinovac Menggunakan
Algoritma Support Vector Machine
(SVM) dan K-Nearest Neighbor (KNN).
Information System Journal
,
4
(2), 42–
46.
Budianto, A., Ariyuana, R., & Maryono, D.
(2019). Perbandingan K-Nearest
Neighbor (KNN) Dan Support Vector
Machine (SVM) Dalam Pengenalan
Karakter Plat Kendaraan Bermotor.
Jurnal Ilmiah Pendidikan Teknik Dan
Kejuruan
,
11
(1), 27–35.
Fadilah, W. R. U., Agfiannisa, D., & Azhar, Y.
(2020). Analisis Prediksi Harga Saham
PT. Telekomunikasi Indonesia
Menggunakan Metode Support
Vector Machine.
Fountain Informatics
J
,
5
(2), 45.
Hakim, I., Nugroho, A., Sukmana, S. H., &
Gata, W. (2020). Sentimen Analisis Stay
Home menggunakan metode
klasifikasi Naive Bayes, Support Vector
Machine, dan k-Nearest Neighbor.
Paradigma-Jurnal Komputer Dan
Informatika
,
22
(2020), 169–174.
Ichwan, M., & Dewi, I. A. (2018). Klasifikasi
Support Vector Machine (SVM) Untuk
Menentukan TingkatKemanisan
Mangga Berdasarkan Fitur Warna.
MIND (Multimedia Artificial Intelligent
Networking Database) Journal
,
3
(2),
16–23.
1988 | Nutrition Classification In Toddlers at UPTD Puskesmas Tigaraksa Using A Comparison of
Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) Methods
Iskandar, J. W., & Nataliani, Y. (2021).
Perbandingan Naïve Bayes, SVM, dan
k-NN untuk Analisis Sentimen Gadget
Berbasis Aspek.
Jurnal RESTI (Rekayasa
Sistem Dan Teknologi Informasi)
,
5
(6),
1120–1126.
Mase, J., Furqon, M. T., & Rahayudi, B.
(2018). Penerapan Algoritme Support
Vector Machine (SVM) Pada
Pengklasifikasian Penyakit Kucing.
Jurnal Pengembangan Teknologi
Informasi Dan Ilmu Komputer E-ISSN
,
2548
, 964X.
Muhammad Yusuf ramadan, D. S. T.
(2019). ”Implementasi Metode
Klasifikasi Support Vector Machine
(SVM) Terhadap Pemakaian Minyak
Goreng.
Jurnal Pengembangan
Teknologi Informasi Dan Ilmu
Komputer
,
3
(2).
Nasution, M. R. A., & Hayaty, M. (2019).
Perbandingan Akurasi dan Waktu
Proses Algoritma K-NN dan SVM
dalam Analisis Sentimen Twitter.
J.
Inform
,
6
(2), 226–235.
Naufal, S. A., Adiwijaya, A., & Astuti, W.
(2020). Analisis Perbandingan
Klasifikasi Support Vector Machine
(SVM) dan K-Nearest Neighbors
(KNN) untuk Deteksi Kanker dengan
Data Microarray.
JURIKOM (Jurnal
Riset Komputer)
,
7
(1), 162–168.
Nikmatun, I. A., & Waspada, I. (2019).
Implementasi Data Mining untuk
Klasifikasi Masa Studi Mahasiswa
Menggunakan Algoritma K-Nearest
Neighbor.
Simetris: Jurnal Teknik
Mesin, Elektro Dan Ilmu Komputer
,
10
(2), 421–432.
Pamungkas, F. S., & Kharisudin, I. (2021).
Analisis Sentimen dengan SVM, NAIVE
BAYES dan KNN untuk Studi
Tanggapan Masyarakat Indonesia
Terhadap Pandemi Covid-19 pada
Media Sosial Twitter.
PRISMA,
Prosiding Seminar Nasional
Matematika
,
4
, 628–634.
Permana, R. A., & Sahara, S. (2019). Metode
Support Vector Machine Sebagai
Penentu Kelulusan Mahasiswa pada
Pembelajaran Elektronik.
Jurnal
Khatulistiwa Informatika
,
7
(1).
Pratama, I. H., & Salamah, U. (2022).
Perbandingan Algoritma K-Nearest
Neighbor Dan Support Vector
Machine Untuk Menentukan Prediksi
Produk-Produk Terlaris Pada Toko
Madura Kecamatan Pondok Aren.
JTIK
(Jurnal Teknik Informatika Kaputama)
,
6
(2), 846–858.
Saidah, S., Adinegara, M. B., Magdalena, R.,
& Caecar, N. K. (2019). Identifikasi
Kualitas Beras Menggunakan Metode
k-Nearest Neighbor dan Support
Vector Machine.
TELKA-Jurnal
Telekomunikasi, Elektronika,
Komputasi Dan Kontrol
,
5
(2), 114–121.
Sugara, B., & Subekti, A. (2019). Penerapan
Support Vector Machine (Svm) Pada
Small Dataset Untuk Deteksi Dini
Gangguan Autisme.
Jurnal Pilar Nusa
Mandiri
,
15
(2), 177–182.
Wati, R. A., Irsyad, H., & Al Rivan, M. E.
(2020). Klasifikasi Pneumonia
Menggunakan Metode Support
Vector Machine.
J. Algoritm
,
1
(1), 21–
32.
©2023 by the authors. Submitted
for possible open-access publication
under the terms and conditions of the Creative
Commons Attribution (CC BY SA) license
(https://creativecommons.org/licenses/by-sa/4.0/).
