Accredited Ranking SINTA 2

Decree of the Director General of Higher Education, Research, and Technology, No. 158/E/KPT/2021
Validity period from Volume 5 Number 2 of 2021 to Volume 10 Number 1 of 2026

Published online on: http://jurnal.iaii.or.id

JURNAL RESTI
(Rekayasa Sistem dan Teknologi Informasi)
Vol. 7 No. 1 (2023) 51 - 57 ISSN Media Electronic: 2580-0760

Classification of Fruits Based on Shape and Color

using Combined Nearest Mean Classifiers
Abdullah1, Agus Harjoko2, Othman Mahmod3
1
Department of Information System, Universitas Islam Indragiri Indonesia
2
Department of Computer Science and Electronics, Universitas Gadjah Mada Indonesia
3
Department of Fundamental and Applied Sciences, Universiti Teknologi PETRONAS, Malaysia
1
[email protected], [email protected], [email protected]

Abstract
Fruit classification is an important task in many agriculture industry. The fruit classification system can be used to identify the
types and prices of fruit. Manual classification of fruit is not efficient for large amount of fruits. The advancement of information
technology has made possible fruit classification be done by a machine. This research aims to propose a fruit classification
methodology based on shape and color. To reduce the effect of lighting variability a color normalization is carried out prior
to feature extraction. The color features used in this research are mean and standard deviation. The shape features are area,
perimeter, and compactness. The classification of an unknown fruit is carried out using the nearest mean classifier. The method
developed in this research is tested using 12 classes of fruits where each class is represented by a number of samples. The
experimental results show that the method proposed in this research provides an accuracy of 95.83% for two samples per class
and 100% for three samples per class. Experiment on small training samples has been conducted to evaluate the performance
of the proposed combined nearest mean classifiers and results obtained showed that the technique was able to provide good
accuracy.
Keywords: fruit classification, nearest mean classifier, color features, shape features

1. Introduction computations more complex. In this paper, a new

simple method of combining multiple classifier namely
Fruit classification which is done manually is inefficient
combined nearest mean classifiers for fruit
and inaccurate for large amount of fruits. The
classification is proposed. In this method, only a few
information technology advancement had made
fruit samples are needed in the training process.
possible the fruit classification be done by a machine or
computer. Fruit classification is meaningful to various This research treated the fruit classification based on
fields i.e. industry, plantation, farming, trading and so shape and color similarity. The classification is
forth [1]. implemented on per fruit item, not the group of fruit.
The different feature used is the color feature (mean and
Combining multiple classifier or ensemble method is
standard deviation of color) and shape feature (area,
considered as a general solution for pattern
perimeter, and compactness). This research result will
classification tasks. Multiple classifier combination
be useful in accelerating the sorting and grading process
goals to achieve the final decision by integrating the
for fruit variants and make easy the fruit trading cost
predictions of several individual classifiers to obtain
decision based on shape (e.g. big or small) and color
comprehensive results. Experimental studies have
(e.g. red or green).
shown that the combination of several classifiers has
been very helpful in improving the classification
2. Research Methods
accuracy [2],[3]
The proposed multiple nearest mean classifier (NMC)
There are several ensemble and deep learning methods
for fruit classification consists of three phases namely
for fruit classification have been proposed [4]-[7].
image preprocessing, feature extraction and classifier
However, deep learning in machine learning uses large
combination as depicted in Figure 1.
amounts of data. Many deep learning algorithms use
multiple layers of neural networks making

Accepted: 21-11-2022 | Received in revised: 16-01-2023 | Published: 02-02-2023

51
 Abdullah, Agus Harjoko, Othman Mahmod
Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi) Vol. 7 No. 1 (2023)

The image of the fruit will be preprocessed to obtain the According to Sen [10],[11] that the classification used
feature of the fruit. Operations such as background as mentioned above is named supervised classification,
subtraction and the normalization of color will be since the class has been noticed and the data sample has
performed on the image of the fruit. Background been available. To develop the supervised
subtraction is performed to separate the image of the classification, earlier a computer system must have the
fruit from its background. Color normalization knowledge that can be developed by learning the
operation is then performed to eliminate the influence sample and recording them in a database [12].
of different lighting.
The fruit classification system follows the structure of
Features in the image of the fruit are extracted and an introduction design system proposed by Yan and
placed in feature vectors. The color features are Gao [13] that includes censor, processing feature
measured by mean and standard deviation on each red, extraction and classifier algorithm. The classification of
green, and blue (RGB) channel. The shape features are fruits is done undirectly by capturing the fruit object’s
measured by area, perimeter and compactness. The area image using the censor. The object’s image that is
of fruit reflects the actual fruit size or weight. The identical with its feature as well the reality is in the same
perimeter of fruit is defined as the area that covers the class [14].
boundary. The compactness of fruit is defined as the
The censor used as the image capture in this system is a
ratio of the area of a fruit to the area of a circle with the
digital camera (or webcam), as shown in Figure 2.
same perimeter.

Figure 2. the hardware scheme of fruit image capture

All the images are captured in a 640 x 480 size. The

images are saved as 24 true color images having 256
grayish levels with RGB color model and bitmap format
file (BMP). Another supporting tool is a tripod to help
capture images at the same distance of 35 cm. The
background color is made same for each image of the
Figure 1. Phases of the proposed methodology fruit object, which is black to avoid the shadow of the
fruit object. The lighting intensity is fixed and not too
The multiple NMC combination consists of three bright to avoid reflection effects on glossy surfaces
nearest mean classifiers [8],[9]. The input features to the which can cause a loss of color information.
first and second classifiers are the color mean and color
standard deviation respectively while the input features According to [15][16] there are several mechanisms
to the third classifier are the area, perimeter, and used in computer vision, one of them is the statistical
compactness. Output from each classifier is the design which uses two phases i.e. training phase and the
similarity value between the unknown object and testing phase’s approach. Generally, the process in the
samples (or training patterns). The similarity value is classification of fruits consists of the main process i.e.
obtained by calculating the euclidean distance between class formation process (training phase) and fruit
the feature vector of the unknown object and the feature classification process (recognition phase).
vector of the sample class mean. The system can be arranged from sub-systems [17]. The
A sample of 84 fruit images that correspond to 12 classification of the fruit system is arranged into two
categories has been used to form the reference values sub-systems which are the class formation system
for each category. The data were divided into a training (SPK), which undertakes the training process, and the
set (43%) and a testing set (57%), with one to three class formation and fruit classification system (SKB),
training samples, were used. All images were 640 x 480 which undertakes the unknown classification of the fruit
pixels with 24-bit true color, 256 levels of gray and, an process into a certain class.
RGB color model. The types of fruits that were used are
limited to variants of apples, mangoes, oranges, pears,
and durian.

DOI: https://doi.org/10.29207/resti.v7i1.4693
Creative Commons Attribution 4.0 International License (CC BY 4.0)
52
 Abdullah, Agus Harjoko, Othman Mahmod
Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi) Vol. 7 No. 1 (2023)

According to Meshram, et. al. [18], the classification of G ( p)

g ( p) = (3)
the fruit architecture system can be seen in the R( p ) + G ( p ) + B( p )
following Figure 3.
B( p )
b( p ) = (4)
R( p ) + G ( p ) + B( p )
with each R (P), G (P), and B (P) as the color intensity
in each component of R (red), G (green), and B (blue)
in pixel p.
The color feature of the fruit image result’s processing
is extracted. The color feature can consist of statistical
data based on the color histogram [21]. The equations
(5), (6) and (7) are used to calculate the color mean, then
equations (8), (9) and (10) are used to calculate the
standard deviation of colors. If the fruit image is x and
the pixel number is p, the mean color of the fruit image
is:

x = (ravg , g avg , bavg )

T
where
P

 r ( p)
p =1
ravg = (5)
P
P

 g ( p)
g avg =
p =1 (6)
Figure 3. The classification of fruit architecture system P
P
2.1. Training phase
 b( p )
p =1
In the training phase, the image of the fruit sample is bavg = (7)
captured through a censor. The fruit sample used P
consists of several fruit samples for each If the fruit image is x and the pixel number is P, the
category/sample afterward the processing is done. deviation standard of the fruit image color is:
The first process is background subtraction to separate T
the fruit from its background by implementing pixel x = (rstd , g std , bstd ) where
subtraction operation. The result is the absolute
 (r ( p) − r )
P
2

subtraction value of the fruit image with its background p =1

avg
(8)
rstd =
[19]. The pixel subtraction operation result is achieved P
from equation (1).
 (g ( p) − g )
P
2

Q(i, j ) = P1 (i, j ) − P2 (i, j )

avg
(1) g std =
p =1 (9)
P

Where P1 is the fruit image and P2 is the background

 (b( p) − b )
P
2

image. The intensity of R, G, and B in the pixel image

bstd =
p =1
avg
()
is the background subtraction result in which the value P
is lesser than the threshold value which is 75 that is Fruit image shape features are also extracted. Fruit
considered as the background. This value is quite ideal image shape features are measured by area,
based on several trials that had been done. circumference, and compactness. The fruit area reflects
The second process is color normalization used to the size and weight of the real fruit. The perimeter is the
disappear the influence of the different lighting [20]. outside part of the object that is side by side with the
The color normalization used equations (2), (3), and (4) pixel background. Perimeter is obtained by calculating
since they are appropriate to the color feature measured the number of pixels in the fruit image borders. If
in each RGB channel. The used equation in doing the x1, x2,..., xn border, the perimeter is measured through
color normalization in each pixel p is:
the equation (11).
R( p )
r ( p) = (2)
R ( p)+ G( p) + B( p)

DOI: https://doi.org/10.29207/resti.v7i1.4693
Creative Commons Attribution 4.0 International License (CC BY 4.0)
53
 Abdullah, Agus Harjoko, Othman Mahmod
Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi) Vol. 7 No. 1 (2023)

n −1 n −1 The similarity measurement is done per groups’ feature

perimeter =  d i =  xi − xi +1 () that can fasten and make simple the query process [26].
i =1 i =1 There are three features which are the color (mean of R,
the distance d1 is equal to 1 to the border with pixel 4 G, and B), colors’ deviation standard (R, G, and B of
line and 1 or √2 to the border with pixel 8 line, this colors’s deviation standard) and, shape (area, perimeter,
research perimeter definition took the pixel 8 line in the and compactness).
outer part of the object and d1 is equal to 1. compactness According to [27][28], the euclidean distance of two
is the fruit shape compact that is measured through the image color vectors mean can be measured. If the mean
equation (12) : of the color fruit image query is stated as vector q and
4. .a the mean class vector as x, the euclidean distance of two
c= () vectors is is shown by the equation (15).
p2
d avg (q, x ) = (r avg q − ravg x ) + (g avg q − g avg x ) + (bavg q − bavg x ) ()
2 2 2

The fruit image shape x is represented by x = (a, p, c)T

where a is the area, p is the perimeter, and c is implementing the same treatment, if the deviation
compactness. standard of fruit color image query is stated as vector q
and mean color in the mean class is vector x, the
The shape and color extraction result of the fruit is kept euclidean distance of two vectors is is shown by the
in the image feature database, SPK creates class mean equation (16).
for each fruit. The class mean or centroid is measured
by using the equation (13). d stdev (q, x ) = (r
stdev q − rstdev x ) + (g stdev q − g stdev x ) + (bstdev q − bstdev x )
2 2 2 ()
1

afterward, if the fruit image query shape is stated as
x= j =1, ni
xi , j () vector q and shape in mean class is vector x, the
ni
euclidean distance of those two vectors is is shown by
xi , j is the jth sample feature vector from class i the equation (17).
where

or create a view from the feature mean measurement in

d bentuk (q, x ) = (a q − a x ) + ( p q − p x ) + (c q − c x ) ()
2 2 2

the image feature database that is classified based on the After the above similarity measurement on each feature
category or class label. group, the similarity measurement is done
2.2. Recognition Phase simultaneously to those three feature groups. The way
is by adding up those three groups’ distances. However,
In the recognition phase, the fruits are classified and the the distance scale in each feature is different, it is
images will be captured through the censor, normalized by subtracting each distance of a certain
implemented preprocessing and feature extraction as in feature with the maximum distance. The normalized
the training phase. distance of each group is around 0-1, so the distance
The extraction result of fruit shape and color is used in similarity total is around 0-3. The equation used to
the image feature query. The classification is done by measure the similarity distance is the equation (18).
measuring the shape and color similarity of the image d (q, x ) d stdev (q, x ) d bentuk (q, x )
d sim (q, x ) =
avg
+ + ()
query with the mean class as equation (13). The max d avg max d stdev max d bentuk
unknown fruit is stated as feature vector q that will be
classified to class i if it is closer to vector mean class i where d sim (q, x ) is the similarity distance,
than others.
max d mean is the maximum distance of the mean color.
The similarity is measured through the vector distance.
Two closest vectors will possess similarities and a little max d stdev is the maximum distance of standard deviation
bit of difference [14]. Generally, the NMC classifier color.
used euclidean distance [22], even according to [23], max d bentuk is the maximum distance in shape’s category.
NMC classifier is also well-known as Nearest Centroid
Classifier [24]. Furthermore, the used distance metric in The similarity distance measurement is done for all
fruit classification is the L2 metric (euclidean metric). mean classes. The classification rule according to [14]
According to Malkov [25] that the euclidean of two is given to two classes w1 and w2. The object’s vectors
vectors x and w is shown by the equation (14). are written as {x1,...,xn}, if x1 is the w1 mean class, the
n
() new object Z is represented in the space as Zx..
x − y =  ( xi − yi )
2

2 i =1 - Classify z to w1 if and if only d2(z, x (1))< d2(z, x (2))

- Classify z to w2 if and if only d2(z, x (2)) < d2(z, x (1))

DOI: https://doi.org/10.29207/resti.v7i1.4693
Creative Commons Attribution 4.0 International License (CC BY 4.0)
54
 Abdullah, Agus Harjoko, Othman Mahmod
Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi) Vol. 7 No. 1 (2023)

The fruit is classified if the minimum value is 3. Results and Discussions

d sim (q, x )   with  = 0.75, vise versa the fruit is There are 12 fruit categories that are made as class
rejected, if the threshold value is 0.75 (scale 0-3) is labels i.e. Fuji Apple, Manalagi Apple, Washington
achieved empirically. The similarity percentage is Apple, Arum Manis A Mango, Arum Manis B Mango,
realtive toward the distance is 100% if the distance is 0. Siam Orange, Honey Mango, Podang Mango, Sunkist
to be classified, the minimum similarity percentage of Orange, Siam Orange, Peer, and Durian. Each class is
fruit is 75% or the similarity distance is 0.75. represented by some of the fruits’ samples. Obtained
feature values a shown in Table 1.
Table 1. Reference Fruit Images Feature Values
Color mean Color standard deviation Shape
Type of fruit
Red Green Blue Red Green Blue Area Perimeter Compactness
Fuji Apple 193.93 136.28 83.73 38.71 50.96 30.03 15911 438 1.04
Manalagi Apple 167.41 180.51 66.67 33.88 34.28 25.70 12581 385 1.07
Wahington Apple 184.34 75.75 64.24 37.20 39.23 30.05 16455 623 0.53
Arum Manis A Mango 131.89 143.69 51.07 27.20 39.23 30.05 31211 1449 0.19
Arum Manis B Mango 112.97 132.85 47.52 17.84 20.79 16.66 17083 920 0.25
Golek Mango 147.24 150.16 36.51 31.02 28.58 27.73 24739 1042 0.32
Honey Mango 105.27 138.55 77.66 22.91 24.81 27.03 20703 1588 0.10
Podang Mango 203.74 143.30 48.70 34.63 34.37 30.31 16436 455 1.00
Sunkist Orange 206.37 114.51 8.17 38.68 37.96 26.97 20846 618 0.69
Siam Orange 176.46 135.07 20.09 36.46. 34.53 33.22 14469 498 0.73
Pear 211.77 191.14 122.68 35.30 41.74 41.80 18324 515 0.87
Durian 117.28 123.81 50.23 19.27 20.72 25.17 72276 10769 0.01

The class formation system’s test is done used 3 ways:

firstly, treated one fruit sample; secondly, treated two
samples and thirdly, treated three samples of fruit
sample for each class. The sample image is taken to
look for its characteristic which is called as training
image.
The class formation system forms the mean class of
each class categorical label as shown in Figure 4 which
is the class formation system’s user face that shows the
mean class formation to three samples in each class.

Figure 5. SKB’s user interface

In evaluating the performance of the proposed method,

48 fruit images from 12 classes were used as the testing
images. The first test used the image of the same fruits,
however, the positions are being changed. The second
test used the image of different fruits that were never
used in the training. The first test result is presented in
Table 2 and the second test is in Table 3.
Table1 shows that the system is able in recognizing and
classifying used for training although the position is
different. Table 2, shows that the system is also able in
Figure 4. SPK’s user interface recognizing and classifying the fruit from the different
samples. Generally, the successfulness level using 2
The fruit classification system’s test is done by doing samples in each class is 95.83. If using 3 samples, the
the image query. This image is called image testing used successfulness level reaches 100%. It means, this
to test the system’s success. The feature of the fruit system has a good reputation.
classification system’s user is viewed in Figure 5.

DOI: https://doi.org/10.29207/resti.v7i1.4693
Creative Commons Attribution 4.0 International License (CC BY 4.0)
55
 Abdullah, Agus Harjoko, Othman Mahmod
Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi) Vol. 7 No. 1 (2023)

Table 2. The Same Fruits Images Test Result [3] M. Astani, M. Hasheminejad, and M. Vaghefi, “A diverse
The Classification ensemble classifier for tomato disease recognition,” Comput.
The image Electron. Agric., vol. 198, p. 107054, 2022, doi:
image
No training Rejected Result %
number
testing True False https://doi.org/10.1016/j.compag.2022.107054.
number [4] J. Kang and J. Gwak, “Ensemble of multi-task deep
1 One 12 11 1 0 91,67 convolutional neural networks using transfer learning for fruit
sample per 24 23 1 0 95,83 freshness classification,” Multimed. Tools Appl., vol. 81, Jul.
class 36 34 2 0 94,44
2022, doi: 10.1007/s11042-021-11282-4.
48 46 2 0 95,83
2 Two 12 11 1 0 91,67 [5] H. S. Gill, O. I. Khalaf, Y. Alotaibi, S. Alghamdi, and F.
samples 24 23 1 0 95,83 Alassery, “Fruit Image Classification Using Deep Learning,”
per class 36 35 1 0 97,22 Comput. Mater. Contin., vol. 71, no. 2, pp. 5135–5150, 2022,
48 46 1 1 95,83 doi: 10.32604/cmc.2022.022809.
3 Three 12 12 0 0 100 [6] C. C. Ukwuoma, Q. Zhiguang, M. B. Bin Heyat, L. Ali, Z.
samples 24 24 0 0 100 Almaspoor, and H. N. Monday, “Recent Advancements in
per class 36 36 0 0 100
Fruit Detection and Classification Using Deep Learning
48 48 0 0 100
Techniques,” Math. Probl. Eng., vol. 2022, p. 9210947, 2022,
doi: 10.1155/2022/9210947.
Table 3. The Other Fruit Image Test Result.
[7] P. Sumari et al., “A Precision Agricultural Application:
Manggis Fruit Classification Using Hybrid Deep Learning,”
The The Classification Rev. d’Intelligence Artif., vol. 35, no. 5, pp. 375–381, 2021,
image image
No
training testing Rejected
Result doi: 10.18280/ria.350503.
True False
number number
% [8] Z. Mai, R. Li, H. Kim, and S. Sanner, “Supervised Contrastive
12 12 0 0 100 Replay: Revisiting the Nearest Class Mean Classifier in Online
One Class-Incremental Continual Learning,” CoRR, vol.
24 24 0 0 100
1 sample
per class
36 36 0 0 100 abs/2103.13885, 2021, [Online]. Available:
48 46 1 1 95,83 https://arxiv.org/abs/2103.13885.
12 12 0 0 100 [9] E. Santucci, “Quantum Minimum Distance Classifier,”
Two
24 24 0 0 100
2 samples Entropy, vol. 19, no. 659, pp. 1–14, 2017, doi:
36 36 0 0 100
per class 10.3390/e19120659.
48 45 2 0 95,83
12 12 0 0 100 [10] P. C. Sen, M. Hajra, and M. Ghosh, “Supervised Classification
Three Algorithms in Machine Learning: A Survey and Review BT -
24 24 0 0 100
3 samples
per class
36 36 0 0 100 Emerging Technology in Modelling and Graphics,” 2020, pp.
48 48 0 0 100 99–111.
[11] K. D. Copsey, Statistical Pattern Recognition. Wiley, 2011.
[12] I. H. Sarker, “Machine Learning: Algorithms, Real-World
4. Conclusion Applications and Research Directions,” SN Comput. Sci., vol.
Classification of fruits using the proposed multiple 2, no. 3, pp. 1–21, 2021, doi: 10.1007/s42979-021-00592-x.
[13] X. Yan and L. Gao, “A feature extraction and classification
nearest mean classifier technique has shown that the algorithm based on improved sparse auto-encoder for round
technique is capable in producing high accuracy with a steel surface defects,” Math. Biosci. Eng., vol. 17, no. 5, pp.
small sample size. The sample number in each class 5369–5394, 2020, doi: 10.3934/MBE.2020290.
influences the system’s ability, the system becomes [14] I. Reppa, K. E. Williams, W. J. Greville, and J. Saunders, “The
relative contribution of shape and colour to object memory,”
better with the sample advancement’s number. Up to 3 Mem. Cognit., vol. 48, no. 8, pp. 1504–1521, 2020, doi:
samples in each class, the system had been able in doing 10.3758/s13421-020-01058-w.
the classification to 48 fruits with 100% in its [15] M. H. Guo et al., “Attention mechanisms in computer vision:
successfulness level or having a good reputation. The A survey,” Comput. Vis. Media, vol. 8, no. 3, pp. 331–368,
2022, doi: 10.1007/s41095-022-0271-y.
image capture process needs to be taken into account, [16] M. Han, H. Wu, Z. Chen, M. Li, and X. Zhang, “A survey of
so the color and shape of the fruit can be represented multi-label classification based on supervised and semi-
well. This way can be applied by using supplemented supervised learning,” Int. J. Mach. Learn. Cybern., 2022, doi:
light and solid-state image censor. In fact, the fruit 10.1007/s13042-022-01658-9.
[17] K. . Kendall and J. . Kendall, System Analysis and Design, 8th
surfaces are not always stainless, sometimes possessing ed. New Jersey: Prentice-Hall, 2020.
stains and dust in which their colors are identical to the [18] V. Meshram, K. Patil, V. Meshram, D. Hanchate, and S. D.
background thus the background subtraction result is Ramkteke, “Machine learning in agriculture domain: A state-
not perfect. It needs an algorithm arrangement and of-art survey,” Artif. Intell. Life Sci., vol. 1, no. September, pp.
1–11, 2021, doi: 10.1016/j.ailsci.2021.100010.
image processing technique to figure out the [19] D. N. Arulnathan, B. C. W. Koay, W. K. Lai, T. K. Ong, and
weaknesses. It needs the fruit classification system’s L. L. Lim, “Background Subtraction for Accurate Palm Oil
hardware so that the fruit sorting and grading process Fruitlet Ripeness Detection,” in 2022 IEEE International
can be done by a machine or robotic system. Conference on Automatic Control and Intelligent Systems
(I2CACIS), 2022, pp. 48–53, doi:
10.1109/I2CACIS54679.2022.9815275.
References [20] N. Phuangsaijai, J. Jakmunee, and S. Kittiwachana,
“Investigation into the predictive performance of colorimetric
[1] H. A. Hambali, S. L. S. Abdullah, N. Jamil, and H. Harun, sensor strips using RGB, CMYK, HSV, and CIELAB coupled
“Fruit classification using neural network model,” J. with various data preprocessing methods: a case study on an
Telecommun. Electron. Comput. Eng., vol. 9, no. 1–2, pp. 43– analysis of water quality parameters,” J. Anal. Sci. Technol.,
46, 2017. vol. 12, no. 1, 2021, doi: 10.1186/s40543-021-00271-9.
[2] O. Sagi and L. Rokach, “Ensemble learning: A survey,” Wiley [21] T. Gevers and A. Smeulders, “Foreword,” Lect. Notes Comput.
Interdisciplinary Reviews: Data Mining and Knowledge Sci. (including Subser. Lect. Notes Artif. Intell. Lect. Notes
Discovery, vol. 8, no. 4. 2018, doi: 10.1002/widm.1249.

DOI: https://doi.org/10.29207/resti.v7i1.4693
Creative Commons Attribution 4.0 International License (CC BY 4.0)
56
 Abdullah, Agus Harjoko, Othman Mahmod
Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi) Vol. 7 No. 1 (2023)

Bioinformatics), vol. 9909 LNCS, p. V, 2016, doi: [25] Y. A. Malkov, “Efficient and robust approximate nearest
10.1007/978-3-319-46493-0. neighbor search using Hierarchical Navigable Small World
[22] N. Ali, D. Neagu, and P. Trundle, “Evaluation of k-nearest graphs,” IEEE Trans. Pattern Anal. Mach. Intell., pp. 31–33,
neighbour classifier performance for heterogeneous data sets,” 2018, [Online]. Available:
SN Appl. Sci., vol. 1, no. 12, pp. 1–15, 2019, doi: https://github.com/nmslib/nmslib%0Ahttp://ann-
10.1007/s42452-019-1356-9. benchmarks.com/hnsw(nmslib).html.
[23] K. Taunk, S. De, S. Verma, and A. Swetapadma, “A Brief [26] R. C. Chen, C. Dewi, S. W. Huang, and R. E. Caraka,
Review of Nearest Neighbor Algorithm for Learning and “Selecting critical features for data classification based on
Classification,” in 2019 International Conference on machine learning methods,” J. Big Data, vol. 7, no. 52, 2020,
Intelligent Computing and Control Systems (ICCS), 2019, pp. doi: 10.1186/s40537-020-00327-4.
1255–1260, doi: 10.1109/ICCS45141.2019.9065747. [27] J. Lv and J. Fang, “A Color Distance Model Based on Visual
[24] S. Johri et al., “Nearest centroid classification on a trapped ion Recognition,” Math. Probl. Eng., vol. 2018, 2018, doi:
quantum computer,” npj Quantum Inf., vol. 7, no. 1, 2021, doi: 10.1155/2018/4652526.
10.1038/s41534-021-00456-5. [28] R. Manisha, “Content Based Image Retrieval using Color and
Texture Feature with Distance Matrices,” Int. J. Sci. Res. Publ.,
vol. 7, no. 8, pp. 512–523, 2012, doi: 10.5121/sipij.2012.3104.

DOI: https://doi.org/10.29207/resti.v7i1.4693
Creative Commons Attribution 4.0 International License (CC BY 4.0)
57