Segmentation of Karhutla Hotspot Point of Indragiri Hilir Regency

2015 and 2016 using Self Organizing Maps (Soms)

Achmad Isya Al Fassa and Ayundyah Kesumawati

Department of Statistics Faculty of Mathematics and Natural Sciences, Islamic University of Indonesia

Keywords: Hotspot Point, Cluster Analysis, Self-Organizing Maps, Disaster Prone Areas.

Abstract: The hotspot point is an indicator to see forest and land fires occurring in the field. Hotspots are usually in the

form of hotspot coordinates. Given the hotspot point is an important role in disaster prevention in Indonesia,

one of which is the disaster of forest and land fires. So, in this case done research to know comparison of

hotspot point and to know disaster prone area at Regional Disaster Management Agency of Indragiri Hilir

Regency - Riau. Cluster analysis using Self Organizing Maps (SOMs) method is one of the methods that can

be used to know the comparison of hotspot points from year to year and to the grouping of disaster prone

areas. The variables used are Hotspot Point 2015, Hotspot Point 2016, Number of Villages, and Area (Km²).

Of all the variables obtained results where Hotspot Point 2015 higher than the Hotspot Point 2016. So that all

the available variables obtained one area that is disaster-prone areas, namely the District Kempas.

1 INTRODUCTION

Indragiri Hilir regency is one of the districts in Riau

Province. Where is Indragiri Hilir regency is one of

the districts that have forest and land large enough

and big, which have disaster problem that is Karhutla

(Fire Forest and Land). Land and Forest Fire

Disasters is one of the major disasters which in the

Regulation of the Minister of Environment and

Forestry of the Republic of Indonesia Number P.32 /

MenLHK / Setjen / Kum.1 / 3/2016 About Forest and

Land Fire Control that Forest and Land Fires are an

event the burning of forests and / or land, either by

nature or by human actions, resulting in

environmental damage that causes ecological,

economic, socio-cultural and political losses. In this

research there are several studies relating to research

conducted by the author, among others : Setiani and

Hakim (2015) "Clustering of Sustainable

Development Indicators in Indonesia Using the

Kohonen Self Organizing Maps Algorithm", Kasih,

Sulhaerati, Septian, Al Fassa, Maulina (2017), in the

research "Cluster Analysis of Polio Immunization

Data in Indonesia in 2016" Using the Self Organizing

Maps Method, Ambarwati and Winarko in (2014).

The purpose of his research is to create an application

system for grouping news articles using the Self

Organizing Maps algorithm, Anis and Isnanto in

(2014). in this study the Self Organizing Maps

method is used to classify geo-referenced data that

integrates the visualization of the output space in

cartographic representation through color

management, and explores the use of the width of the

border line between elements with geographic

references, calculated according to the best distance

in the input space between locations, and Akbar

(2016) in a research report on practical work at the

Jakarta Capital Region Regional Disaster

Management Agency with the title "Cluster Analysis

Using Self Organizing Maps Method for Grouping

Fire-Prone Hazardous Areas Based on Types of

Damaged Facilities in DKI Jakarta Province 2013-

2015".

2 LITERATURE REVIEW

2.1 Disaster

According to the causes, disasters can be divided into

(1) Natural Disasters due to natural factors, (2) Non-

Natural Disasters due to non-environmental factors,

and (3) Social Disasters due to human factors.

(Sarwidi, 2015).

336

Fassa, A. and Kesumawati, A.

Segmentation of Karhutla Hotspot Point of Indragiri Hilir Regency 2015 and 2016 Using Self Organizing Maps (SOMs).

DOI: 10.5220/0008521603360341

In Proceedings of the International Conference on Mathematics and Islam (ICMIs 2018), pages 336-341

ISBN: 978-989-758-407-7

2.2 Forest and Land Fires

Forest and Land Fires, hereinafter referred to as

Karhutla, are an event of burning of forests and / or

land, both natural and by human acts, thus causing

ecological, economic, socio-cultural and political

losses. (Number P.32 / MenLHK / Setjen / Kum.1 /

3/2016 Article 1).

2.3 Descriptive Statistics Analysis

Descriptive Statistics are methods or ways used to

summarize and present data in the form of tables,

graphs or numerical summary of data (Hasan, 2001).

Explain that descriptive statistics are part of statistics

that studies how data collection and data presentation

are so easy to understand. Descriptive statistics only

deal with the matter of deciphering or providing

information about a data or state. With the word

descriptive statistics function explain the state,

symptoms, or problems.

2.4 Cluster Analysis with Self

Organizing Maps

According to (Han et al, 2011) Clustering is the

process of grouping the data set into several groups

so that objects within a group have many similarities

and have many differences with other group objects.

Differences and their similarities are usually based on

the attribute value of the object and can also be a

distance calculation. Clustering itself is also called

Unsupervised Classification, because clustering is

more to be learned and noticed. Cluster analysis is the

process of partitioning a set of data objects into

subsets. Each subset is a cluster, so objects that are in

a cluster are similar to each other, and have

differences with objects from other clusters.

Partitions are not done manually but with clustering

algorithms. Therefore, Clustering is very useful and

can find unknown groups in the data. There are

several approaches used in developing clustering

methods. The two main approaches are Clustering

with Partitioning and Clustering approaches with a

hierarchical approach. Clustering with partitioning

approach or often referred to as partition-based

clustering grouping data by sorting through the data

that is analyzed into existing clusters. Clustering with

hierarchical approach or often called hierarchical

clustering classifies data by creating a hierarchy of

dendrograms where similar data will be placed in

adjacent hierarchies and not in a remote hierarchy. In

addition to both approaches, there is also Clustering

with automatic mapping approach (Self Organizing

Maps / SOM). Self Organizing Map is a leading

method of neural network approach for clustering,

after competitive learning (Han & Kamber 2001).

SOM is different from competitive learning that is the

nerve in a learning environment to recognize the

environmental part of the input space. SOM

recognizes the distribution (such as competitive

learning) and topology of input vector through

training process (Demuth & Beale 2003) SOM shows

three characteristics: the competition is each vector of

weights competing each other to be the winning node,

the cooperative ie each winning node in cooperation

with its environment, and adaptation of the winner's

node change and environment (Larose 2004).

3 METHODS

This research was conducted while carrying out

practical work on January 16 until February 16, 2017

at the Regional Disaster Management Agency

Indragiri Hilir.

3.1 Data Source

The data obtained by the researcher is secondary data

taken from documents / archives of Regional Disaster

Management Agency of Indragiri Hilir Regency and

Central Statistics Agency of Indragiri Hilir Regency.

The variables in this study are variables

Hotspot Point 2015, Hotspot Point 2016, Number

of Villages, and Area (Km²).

3.2 Step Analysis

The method used in this research is descriptive and

comparative statistics to determine the general

description of the existing data variables, while for

grouping the existing variable used Cluster Analysis.

Where Cluster Analysis is used to identify areas

prone to forest and forest land (Karhutla). The

software used in this analysis is SPSS, Mic. Excel, R,

and Software Support.

3.3 Self-Organizing Maps

According to experts (Han et al, 2011) Clustering is

the process of grouping data into several groups so

that objects in one group have many similarities and

have many differences with objects in other groups.

Segmentation of Karhutla Hotspot Point of Indragiri Hilir Regency 2015 and 2016 Using Self Organizing Maps (SOMs)

337

Figure 1: Architecture of Self Organizing Maps.

In cluster analysis there is a measure of distance

where distance measurements are applied to metric

scale data. Actually, it is a measure of dissimilarity,

where large distances show little similarity whereas

short / small distances indicate that an object is more

similar to another object. The difference with the size

of the correlation is that the size of the distance focus

is on the magnitude of the value. The most commonly

used size is the Euklidian distance. Euklidian distance

is the magnitude of the distance of a straight line that

connects objects. Suppose there are two objects,

namely A with coordinates (x) and B with coordinates

(j), the distance between the two objects can be

measured by the formula :









  









D(w j , xn ) = Σk=i (wij − xni )

(1)

Description :

D (j) = Distance between object i

and object j

xni = The object value of the

variable i to k

w = The object value of the variable

i to k

 = Many variables

observed

4 RESULTS AND DISCUSSION

Figure 2: Karhutla Hotspot Point Chart 2015-2016.

In figure 2 above is a bar graph hotspot point

karhutla 2015-2016. Where the graph shows the

number of hotspots of forest and land fires occurring

in all districts in Indragiri Hilir Regency. Where in

figure 2 hotspot areas that have the highest point is

the District Kempas.

Figure 3: Descriptive Statistics Analysis of Karhutla

Hotspot Point Data Year 2015 and 2016.

From Figure 3 during 2015 the value of N Valid =

20 and N Missing = 0. This shows that valid data in

2015 has 20 data and no missing data. Mean Value =

18.55, Median Value = 5.50, Minimum Value = 0,

Maximum Value = 111, and

Sum Value = 371. While in 2016 the value of N

Valid and N Missing is the same as 2015, N Valid =

20 and N Missing = 0 it also shows that valid data in

2016 has 20 data and no missing data. Mean Value =

1.30, Median Value = .00, Minimum Value = 0,

Maximum Value = 6, and Sum Value = 26.

Figure 4: Histogram Hotspot Point 2015.

Figure 5: Histogram Hotspot Point 2016.

ICMIs 2018 - International Conference on Mathematics and Islam

338

From Figure 4 and Figure 5 is the output of

descriptive statistical analysis using SPSS software.

Where from Figure 4 and Figure 5 it can be concluded

that the data used by researchers, be it data hotspot

point 2015 and 2016 hotspot data point is abnormal

data.

After the descriptive statistical analysis, then in this

cluster analysis, researchers used multiple variables,

namely: District, Point 2015 and Point Hotspot 2016

Number of Rural and Regional Area (km2). The

results of the cluster analysis performed by the

researchers is to divide or classify subdistricts in

Indragiri Hilir into several groups: high, medium,

light, and very light, according to the number of

points hotspots owned by each district.

Figure 6: Training Progress Iterate.

The results of the training progress output are the

results showing the number of iterations against the

average distance to the nearest unit. It can be seen that

iteration shows the convergence from the 60th

iteration. Based on this training progress chart it can

be seen that iteration can be done 100 times iteration.

The process in Self Organizing Maps creates a Self-

Organizing Maps model and, in the process, using R

application will generate a diagram containing

several circles that will be closely topological if the

characteristics are the same.

Figure 7 is the result of cluster analysis above shows

the result of clustering previously explained by

researchers, in figure 7 above can be seen the

researcher make Venn diagram by using hexagonal

display with grid 4 x 5. This diagram is based on the

data input then processed using Kohonen algorithm

by using 4 variables.

Figure 7: Output Clustering Self Organizing Maps.

Once formed Venn diagram can be seen the identity

of each circle contained from the output obtained, in

each circle there are identities of each variable are:

green color is a hotspot point in 2015, yellow color is

a hotspot point 2016, cream color is the number of

villages, and the white color is the area, with the

description as follows:

1. Cluster I: This cluster is green; this cluster

shows as a vulnerable area or high group.

2. Cluster II: This cluster is blue; this cluster shows

the moderate category.

3. Cluster III: This cluster is orange; this cluster

shows the category of light category.

4. Cluster IV: This cluster is red; this cluster shows

as a very lightweight group.

In this cluster analysis, cluster number of 20 clusters

is obtained based on the number of sub-districts in

Indragiri Hilir Regency. There are 4 groups in the

cluster analysis conducted by the researcher that is:

high, medium, light, and very light. This group is

based on the number of hotspots that exist in each

sub-district in Indragiri Hilir district with 2015

hotspot variables, hotspots 2016, number of villages,

and total area.

Segmentation of Karhutla Hotspot Point of Indragiri Hilir Regency 2015 and 2016 Using Self Organizing Maps (SOMs)

339

Figure 8: Results Cluster Analysis with Self Organizing

Maps.

From Figure 8 is the result of clustering Cluster

analysis using Self Organizing Maps Method. Where

in group 1 is a group that is categorized by high

category, group 2 is categorized by medium category,

group 3 is categorized by light category, group 4 is

categorized with very light category. From the results

of the grouping the researchers get a conclusion that

group 1 is categorized by high category is the area of

disaster-prone forest fire and land.

5 CONCLUSIONS

Based on the results of comparison, descriptive

analysis and discussion Cluster Analysis, in the case

study, it can be concluded that:

1. The condition of forest and land fires in

Indragiri Hilir Regency in 2015 until 2016 has

decreased the number of fire incidents. This is

due to the hotspot point data obtained in 2016

less than 2015, according to the comparison

results described by previous researchers.

2. Descriptive Analysis Results in 2015 N values

Valid = 20 and N Missing = 0 this shows that

valid data in 2015 there are 20 data and do not

have missing data. Mean Value = 18.55, Median

Value = 5.50, Minimum Value = 0, Maximum

Value = 111, and Sum Value = 371. While in

2016 the value of N Valid and N Missing is

equal to 2015, N Valid = 20 and N Missing = 0

It also shows that valid data in 2016 has 20 data

and no missing data. Mean Value = 1.30,

Median Value = .00, Minimum Value = 0,

Maximum Value = 6, and Sum Value = 26.

3. Grouping is done by Self Organizing Maps

method that produces cluster. Where cluster I is

an area that has high hotspot point of 1

subdistrict, cluster II is an area that has hotspot

point which consist of 8 districts, cluster III is

area having hotspot point which consist of 5

districts, and cluster IV is area having a very

light hotspot point that consists of 6 districts.

Cluster I has members of Kempas Sub-district.

Cluster II consists of Bird Island Sub-District,

Keritang Sub-District, Tempuling Sub-District,

Gaung Sub-District, Kemuning Sub-District,

Mandah Sub-District, Gaung Anak Serka Sub-

District, Pelangiran Sub-District. Cluster III

consists of District Tembilahan, District Enok,

District Reteh, District Kateman, Belengkong

Bay District. Cluster IV consists of Batang

Tuaka Subdistrict, Concong Sub-District, Kuala

Indragiri Sub-District, Sungai Batang

Subdistrict, Tembilahan Hulu Sub-District,

Tanah Merah Sub-district.

REFERENCES

Akbar, Purnama. 2016. Analisis Cluster Menggunakan

Metode Self Organizing Maps (SOM) Untuk

Pengelompokan Daerah Rawan Bencana Kebakaran

Berdasarkan Jenis Sarana yang Rusak di Provinsi DKI

Jakarta Tahun 2013-2015. Laporan kerja praktek tidak

diterbitkan, Jurusan Statistika, Fakultas MIPA

Universitas Islam Indonesia, Yogyakarta.

Demuth H, Beale M. 2003. Neural Network Toolbox for

Use with MATLAB. USA: The MathWorks, Inc.

Hasan, I. 2001. Pokok-Pokok Materi Statstik 2. Jakarta:

Bumi Aksara

Han, J., Kamber, M., Jian, P. 2011. Data Mining: Concepts

and Techniques, 3rd ed. Morgan Kaufmann.

Han J, Kamber M. 2001. Data Mining: Concepts and

Techniques. USA: Academic Press.

Larose DT. 2004. Descovering Knowladge in Data: An

Introduction to Data Mining. USA: John Wiley&Sons

Inc.

Liputan6.com/read/2322792/sisa-kebakaran-hutan-di-riau-

kembali-membara

Peraturan Kepala Badan Nasional Penanggulangan

Bencana Nomor 3 Tahun 2008 tentang BNPB.

Peraturan Kepala Badan Nasional Penanggulangan

Bencana Nomor 3 Tahun 2008 tentang BPBD.

Peraturan Kepala Badan Nasional Penanggulangan

Bencana Nomor 9 Tahun 2008 tentang TRC (Tim

Reaksi Cepat).

Peraturan Menteri Lingkungan Hidup dan Kehutanan.

Nomor P.32/MenLHK/Setjen/Kum.1/3/2016 Pasal 1.

Tentang Kebakaran Hutan dan Lahan (Karhutla).

Peraturan Menteri Lingkungan Hidup dan Kehutanan.

Nomor P.32/MenLHK/Setjen/Kum.1/3/2016 Pasal 1.

Tentang Hutan.

Peraturan Menteri Lingkungan Hidup dan Kehutanan.

Nomor P.32/MenLHK/Setjen/Kum.1/3/2016 Pasal 1.

Tentang Lahan.

Sarwidi. 2015. Pengetahuan Dasar Kebencanaan dan

Kegempaan. Universitas Islam Indonesia. Yogyakarta.

Statistikolahdata.com/2010/10/analisis-perbandingan.html

ICMIs 2018 - International Conference on Mathematics and Islam

340

Sipongi.menlhk.go.id/hotspot/luas_

Sigama.web.id/index.php/kebakaran-hutan.html

Tnrawku.wordpress.com/category/kebakaran-hutan-dan-

lahan/

Undang-Undang Nomor 26 Tahun 2007 tentang Wilayah.

Yulianto, Safa’at., K.H. Hidayatullah. 2014. Analisis

Klaster Untuk Pengelompokan Kabupaten/Kota di

Provinsi Jawa Tengah Berdasarkan Indikator

Kesejahteraan Rakyat. Akd. Statistika Muhammadiyah

Semarang

Segmentation of Karhutla Hotspot Point of Indragiri Hilir Regency 2015 and 2016 Using Self Organizing Maps (SOMs)

341