TELKOMNIKA, Vol.16, No.3, June 2018, pp.

1351~1357
ISSN: 1693-6930, accredited A by DIKTI, Decree No: 58/DIKTI/Kep/2013
DOI: 10.12928/TELKOMNIKA.v16i3.8910  1351

K Means Clustering and Meanshift Analysis for

Grouping the Data of Coal Term in Puslitbang tekMIRA
1 2 3
Rolly Maulana Awangga* , Syafrial Fachri Pane , Khaera Tunnisa ,
4
Iping Supriana Suwardi
1,2,3
Program in Informatics Engineering, Politeknik Pos Indonesia , Indonesia
4
School of Electrical Engineering and Informatics, Bandung Institute of Technology, Indonesia
*Corresponding author, e-mail: awangga@poltekpos.ac.id

Abstract
Indonesian government agencies under the Ministry of Energy and Mineral Resources have
prob lems in classifying data dictionary of coal. This research conduct grouping coal dictionary using K-
Means and MeanShift algorithm. K-means algorithm is used to get cluster value on character and word
criteria. The last iteration of Euclidian distance calculation data on k -means comb ine with Meanshift
algorithm. The meanshift calculates centroid b y selecting different b andwidths. The result of grouping
using k-means and meanshift algorithm shows different centroid to find optimum b andwidth value. The
data dictionary of this research has sorted in alphab etically.

Keywords: coal dictionary, clustering, K-means, meanshift

Copyright © 2018 Universitas Ahmad Dahlan. All rights reserved.

1. Introduction
Data Mining is a process of extracting data or filtering data by utilising a collection of a
big data. The data through a series of operations to obtain valuable information from the dat a.
According to Daryl Pregibon, the composition of artificial statistics, artificial intelligence, and
database research called data mining [1]. Data mining techniques used to explore the data
using classification, prediction, grouping, outlier detection, association rules, sequence analysis,
time series analysis and text mining, and some new techniques such as social networking
analysis and sentiment analysis [2]. The center of research and development mineral and coal
technology (Puslitbang tekMIRA) is an Indonesia government institution under the Ministry of
Energy and Mineral Resources. Puslitbang tekMIRA has a dictionary of coal. This coal
dictionary has one thousand coal terms. It takes time between three to five minutes to find a
term inside a coal dictionary.
Based on these problems, data mining method is used to classify data dictionary. This
research conduct grouping data dictionary of a coal term using data mining method. K-Means
and Meanshift Algorithm were chosen in this research. K-Means algorithm was used to
categories students with skills such as cognitive, communication and relational[3], to evaluate
student achievement levels for course content [4], and to group data based on user information
created on SNS and recommend to users in the future [5]. The grouping data based on user
sentences by utilising regularity among data pursued by the user using K -Means Algorithm [6].
The genetic algorithm and K-Means are used to calculate clustered centroid with heterogeneous
populations that lead to better results than using random numbers [7]. The Meanshift algorithm
is used to accurately detect the location of target tracking [8] by using thi s method and then
make it easy the accuracy of the calculation of the tracking results [9]. The Meanshift algorithm
is also used to solve facial-detection and tracking-based systems [10], if an imbalance occurs, it
will affect the performance of [11].
In this research the algorithm grouping a coal term in a dictionary based on character
and word in a cluster. The cluster serve data dictionary of coal based on predetermined criteria.
The result data of clustering using K-Means and Meanshift algorithm is shown using the
matloplib plot. Matplotlib is a Python package for Plotting that produces quality production
graphs. Matplotlib is designed to be able to create simple and complex plots with multiple
commands [12].

Received September 18, 2017; Revised February 8, 2018; Accepted April 5, 2018
1352  ISSN: 1693-6930

2. Research Method
2.1. K-Means
K-Means Clustering is a grouping of data, where the data in K Means Clustering K is
the amount of data or the number of constants. Means is the average value of the data set as
Cluster [7]. So K-Means is a method to analyse data or called data mining method where this
data modelling method without using supervise or unsupervised method and K -Means is a
method to classify data by using partition system. The K-Means method solves large amounts of
data in groups, and the data has the same characteristics as ot her data. Also, the team feature
also has a feature [13]. Similitude is some matrix used for similarities between instances in a
cluster. K-Means is an algorithm used to generate k clusters from a collection of data sets in a
simple way [14]. Algorithm 1 is an explanation of k-means implementation. Algorithm 1: k-
means clustering: Randomly select k cluster centers c1, ... ck, Repeat, Set each data entity to
the closest cluster center ci, Change the cluster center with the average cluster i, until the
cluster center does not change[15]. K - Means algorithm formula d = distance, j = amount of
data, c = centroid, x = data, c = centroid 1 . The Euclidean distance formula is described in
Equation 2.

( ) √∑

D (i, j) is the distance of data between i and center cluster j. Xk i is the data to i on
attribute data to k. Xkj is the center point to j at attribute to k.
Recalculate cluster center with current cluster membership. The cluster center is the
average of all data/objects in a particular cluster. If desired it can also use the median of the
cluster. So mean (mean) is not the only size that can use.Reassign each object using the new
cluster center. If the cluster center does not change again, then the clustering process is
complete. Alternatively, return to step number 3 until the center of the cluster does not change
anymore [16].

2.2. Meanshift
The average shift method is a method for determining the maximum function limit of
density with separate data from a function. The average shift use as a media grouping method,
in which each mode represents each group [17]-18]. The average shift method classifies data in
its search mode that directs and moves data to the point region along with the iteration of the
data environment built with the Gaussian Kernel [19]. Starting from one data point and iteratively
will improve the approximate mode [20]. The Gaussian kernel is a differentiated multivariate
kernel function used for actual calculations in assumptions [21]. Bandwidth is a free parameter
that shows the effect on the estimated density generated.

3. Results and Analysis

Grouping of data of coal dictionary using K-Means and Meanshift Algorithm. The
grouping results are displayed using matplotlib plot. The sample data taken are vowel data (A, I,
U, E, O) for the coal dictionary. Calculation data using K-Means algorithm. The result data of the
last iteration, i.e. data that has been stable and without changes.

3.1. Result K-Means

3.1.1. The letter A
The data of the last iteration calculation using euclidian distance k- means for letter A.
Clustering data using character and word criteria and use the centroid value of the highest and
lowest values on the criteria. In table 1 is a cluster value that has a fixed and unchanged cluster.

TELKOMNIKA Vol. 16, No. 3, June 2018: 1351-1357

TELKOMNIKA ISSN: 1693-6930 1353

Table 1. Data cluster of the letter A

Character Word
19.04536 3.365126
12.0041 3.78689
7.033561 8.84361
17.00289 1.21311
....... .......
8.006149 7.78689
11.00447 4.78689
27.20259 11.60749

3.1.2. The letter I

The data of the last iteration calculation using euclidian distance k - means for letter I.
Clustering data using character and word criteria and use the centroid value of the highest and
lowest values on the criteria. In Table 2 is a cluster value that has a fixed and unchanged
cluster.

Table 2. Data cluster of the letter I

Character Word
56.4358 1.234823
62.39391 4.774678
37.21559 20.42897
43.18565 14.5126
....... .......
59.4138 1.777967
67.47592 9.838576

3.1.3. The letter U

The data of the last iteration calculation using euclidian distance k - means for letter U.
Clustering data using character and word criteria and use the centroid value of the highest and
lowest values on the criteria. In Table 3 is a cluster value that has a fixed and unchanged
cluster.

Table 3. Data cluster of the letter U

Character Word

0 6.297319

24.02082 17.96443
24.02082 17.96443

....... .......
29.01724 22.93403
23.02173 16.97265

3.1.4. The letter E

The data of the last iteration calculation using euclidian distance k- means for letter E.
Clustering data using character and word criteria and use the centroid value of the highest and

K Means Clustering and Meanshift Analysis for Grouping… (Rolly Maulana Awangga)
1354  ISSN: 1693-6930

lowest values on the criteria. In Table 4 is a cluster value that has a fixed and unchanged
cluster.

Table 4. Data Cluster of the Letter E

Character Word

13.15295 1.009718

14.14214 1.973457

14.14214 1.973457
....... .......
7.071068 5.100934
10 2.625744

3.1.5. The letter O

The data of the last iteration calculation using euclidian distance k - means for letter O.
Clustering data using character and word criteria and use the centroid value of the highest and
lowest values on the criteria. In Table 5 is a cluster value that has a fixed and unchanged
cluster.

Table 5. Data Cluster of the Letter O

Character Word

12.33333 6.342096

3.33333 2.687423
....... .......
2.33333 3.726783
14.66667 20.73377

3.2. Meanshift
3.2.1. The letter A
In Figure 1 is the plot result for the data of the letter A. The data used is the term data of
coal letter A. The value used by the data of the term A to get bandwidth one on the plot
meanshift is at point 6.76.

Figure 1. Plot Mean Shift letter A

TELKOMNIKA Vol. 16, No. 3, June 2018: 1351-1357

TELKOMNIKA ISSN: 1693-6930 1355

3.2.2. The letter I

In Figure 2 is the plot result for the data of the letter I. The data used is the term data of
coal letter I. The value used by the data of the term I to get bandwidth one on the plot meanshift
is at point 18.95.

Figure 2. Plot MeanShift letter I

3.2.3. The letter U

In Figure 3 is the plot result for the data of the letter U. The data used is the term data of
coal letter U. The value used by the data of the term I to get bandwidth one on the plot
meanshift is at point 5.75.

Figure 3. Plot MeanShift letter U

3.2.4. The letter E

In Figure 4 is the plot result for the data of the letter E. The data used is the term data
of coal letter E. The value used by the data of the term I to get bandwidth one on the plot
meanshift is at point 3.36.

K Means Clustering and Meanshift Analysis for Grouping… (Rolly Maulana Awangga)
1356  ISSN: 1693-6930

Figure 4. Plot MeanShift letter E

3.2.5. The letter O

In Figure 5 is the plot result for the data of the letter O. The data used is the term data
of coal letter O. The value used by the data of the term I to get bandwidth one on the plot
meanshift is at point 7.

Figure 5. Plot MeanShift letter O

4. Conclusion
Grouping a coal dictionary using the k-means algorithm produces a cluster value on
character and word criteria. The last iteration of the Euclidian distance calculation results in
different cluster values in each alphabet. The centroid value of the k -means calculation is then
combined with the MeanShift algorithm. Centroid calculation results of the meanshift algorithm
result in different bandwidth. Letter A produces a bandwidth of 6.76, for the letter I produce
bandwidth 18.95, letter U produces bandwidth 5.75, and letter E produces bandwidth 3.36. The
grouping of coal dictionaries in this study facilitates for a group of terms in the letter.
Suggestions for future research, can classify the data dictionary of coal for the second, third and
subsequent letters, and combine other algorithms in grouping such as K-Nearest Neighbour
(KNN) algorithm.

TELKOMNIKA Vol. 16, No. 3, June 2018: 1351-1357

TELKOMNIKA ISSN: 1693-6930 1357

References
[1] F Gorunescu, Data Mining: Concepts, models and techniques. Springer Science & Business Media,
2011; 12.
[2] MI Ramadhan et al. An analysis of natural disaster data b y using k-means and k-medoids algorithm
of data mining techniques. in Quality in Research (QiR): International Symposium on Electrical and
Computer Engineering, 2017 15th International Conference on. IEEE. 2017: 221–225.
[3] J Yi, S Li, M Wu, HA Yeung, WW Fok, Y Wang, F Liu. Cloud-b ased educational b ig data application
of apriori algorithm and k -means clustering algorithm b ased on students’ information. in Big Data and
Cloud Computing (BdCloud), 2014 IEEE Fourth International Conference on. IEEE. 2014: 151–158.
[4] L Guoli, W Tingting, Y Limei, L Yanping, G Jinqiao. The improved research on k -means clustering
algorithm in initial values. in Mechatronic Sciences, Electric Engineering and Computer (MEC),
Proceedings 2013 International Conference on. IEEE. 2013; 2124–2127.
[5] SH Jung, JC Kim, CB Sim. Prediction data processing scheme using an artificial neural network an d
data clustering for big data. International Journal of Electrical and Computer Engineering (IJECE).
2016; 6(1): 330-336.
[6] A Bedboudi, C Bouras, MT Kimour. An heterogeneous population-based genetic algorithm for data
clustering. Indonesian Journal of Electrical Engineering and Informatics (IJEEI). 2017; 5(3): 275–284.
[7] Q Shi, L Xu, Z Shi, Y Chen, Y Shao. Analysis and research of the campus network user’s b ehavior
b ased on k -means clustering algorithm . in Digital Manufacturing and Automation (ICDMA), 2013
Fourth International Conference on. IEEE, 2013: 196–201.
[8] Y Cheng, W Hongyu, W Xiaohong. Positioning method research for unmanned aerial vehicles b ased
on meanshift tracking algorithm . in Control and Decision Conference (CCDC), 2017 29th Chinese.
IEEE 2017: 989–994.
[9] RM Awangga, NS Fathonah, TI Hasanudin. Colenak: GPS tracking model for post-stroke
rehab ilitation program using AES-CBC URL encryption and QR-Code. in 2017 2nd International
conferences on Information Technology, Information Systems and Electrical Engineering (ICITISEE).
IEEE. Nov 2017: 255–260. [Online]. Available: http://ieeexplore.ieee.org/document/8285506/
[10] A Salhi, Y Moresly, F Ghozzi, A Fakhfakh. Face detection and tracking system with b lock -matching,
meanshift and camshift algorithms and kalman filter. in Sciences and Techniques of Automatic
Control and Computer Engineering (STA), 2017 18th International Conference on. IEEE. 2017: 139–
145.
[11] IN Yulita, S Purwani, R Rosadi, RM Awangga. A quantization of deep b elief networks for long short-
term memory in sleep stage detection. in Advanced Informatics, Concepts, Theory, and Applications
(ICAICTA), 2017 International Conference on. IEEE, 2017: 1-5.
[12] N Ari, M Ustazhanov. Matplotlib in python. in 2014 11th International Conference on Electronics,
Computer and Computation (ICECCO). IEEE, sep 2014. [Online]. Available:
https://doi.org/10.1109/icecco.2014.6997585
[13] J MacQueen et al. Some methods for classification and analysis of multivariate ob servations in
Proceedings of the fifth Berkeley symposium on mathematical statistics and probability. Oakland, CA,
USA. 1967; 1(14): 281–297.
[14] AK Jain. Data clustering: 50 years beyond k-means. Pattern recognition letters. 2010; 31(8): 651-666.
[15] Z Gheid, Y Challal. Efficient and privacy-preserving k-means clustering for big data mining. in
Trustcom/BigDataSE/I SPA, IEEE. 2016: 791–798.
[16] K Obbie. Penerapan algoritma klasifikasi data mining id3 untuk menentukan penjurusan siswa sman6
semarang. Skripsi, Fakultas Ilmu Komputer, 2014.
[17] Y Liu, SZ Li, W Wu, R Huang. Dynamics of a mean-shift-like algorithm and its applications on
clustering. Information Processing Letters. 2013; 113(1): 8–16.
[18] G Chen, Q Chen, D Zhang. Mean shift: A method for measurement matrix of compressive sensing . in
Digital Home (ICDH), 2014 5th International Conference on. IEEE. 2014: 64–69.
[19] YA Ghassabeh. A sufficient condition for the convergence of the mean shift algorithm with gaussian
kernel. Journal of Multivariate Analysis. 2015; 135: 1–10.
[20] K Fukunaga, L Hostetler. The estimation of the gradient of a density function, with applications in
pattern recognition. IEEE Transactions on information theory. 1975; 21(1): 32–40.
[21] C Yang, R Duraiswami, NA Gumerov, L Davis. Improved fast gauss transform and efficient kernel
density estimation. in null. IEEE. 2003: 464.

K Means Clustering and Meanshift Analysis for Grouping… (Rolly Maulana Awangga)