Publication Repository ICT2015 tati
Determining Citizen Complaints to The Appropriate
Government Departments using KNN Algorithm
Suhatati Tjandra*,
Amelia Alexandra Putri Warsito**
Department of Informatics Engineering
Sekolah Tinggi Teknik Surabaya
Surabaya, Indonesia
*[email protected], **[email protected]
Abstract—Participation of citizens in the process of city
development is very important. To achieve good governance and
democratic, the citizen can participate by providing complaints,
information, or advices. In the current system, complaints are
handled manually by 1-2 operators, whereas speed and accuracy
are needed. The problem is this manual handling causes errors in
the determination of appropriate government departments that
handle the complaint. This research will propose a system that
aims to determine the appropriate government department with
complaints given by the citizen with the implementation of KNearest Neighbor (KNN) algorithm, to reduce human errors.
This algorithm is one of text classification algorithms, which in
this research, is used to classify complaints which the texts in
Indonesian language. The input of the system is complaint given
by the citizen and the output is the name of the appropriate
government department, which is in accordance with the contents
of the complaint.
Keywords—text classification, k-nearest neighbor, supervised
learning, indonesian language, complaint
I.
INTRODUCTION
The growth of information technology that is rapidly
increasing makes the elements in it are affected. This
information technology greatly affects the processing,
distribution, keeping and communicating of information.
Existing institutions, such as companies, organizations, and
governments require information in the development of
institution itself, especially information that can support the
progress of an institution. However, to communicate this
information is sometimes not supported by good infrastructure,
so that there is no communication running smoothly.
Citizen likes to give complaints against what has happened
to the related institution. Citizen does not only convey a
complaint, but there also conveyed appreciation for what
happened. To accommodate complaints from the citizen, there
are institutions that provide the facilities to convey complaints
through various communication media. However, the
submitted complaints are sometimes not given a response and
there is a given response, but the response is given in a long
time. This happens because the institution is quite difficult to
determine the appropriate department in accordance with the
content of the complaint, because the relevant departments
should be determined manually and sometimes complaints
Judi Prajetno Sugiono
Department of Industrial Engineering
Sekolah Tinggi Teknik Surabaya
Surabaya, Indonesia
[email protected]
given to the departments which are not appropriate. This causes
responses given in a long time.
Based on the explanation, in this research, we propose a
system that can be used to handle these problems. A
classification algorithm in machine learning will be
implemented in this system. Arthur Samuel, who is a pioneer in
the field of machine learning, in 1959, defines that machine
learning as the field of study that gives computers the ability to
learn without being explicitly programmed [1]. One of the
categories of machine learning is a classification. In this
research, classification is conducted on the text, so it is called
as text classification or text categorization. The goal of the text
classification is to classify documents into a number of
predefined categories. There are many algorithms have been
developed for text classification, such as K-Nearest Neighbors
(KNN), Naïve Bayes, Decision Tree, Support Vector Machine
(SVM), etc [2].
Among all these algorithms, KNN is a widely used text
classifier because of its simplicity and efficiency [3]. Many
researches have shown that the KNN algorithm produces very
good performance in the experiments. In the research of short
text classification, KNN algorithm shows a better performance
compared with the Naïve Bayes and SVM algorithm [4].
Moreover, the other research, that uses Reuters corpus as a
dataset, shows that the KNN algorithm also produces very good
performance, when compared with the Naïve Bayes, Rocchio,
and C4.5 algorithm [5]. Based on that, we propose a system
that implements KNN algorithm to classify the text of the
complaint into a number of the government departments.
In this section, we explain the background and purpose of
this research. In the second section, we explain the proposed
system along with the input and output of the system. A
detailed explanation of KNN algorithm to determine the
appropriate government departments in accordance with the
contents of the complaint will be outlined in the third section.
The fourth and fifth sections are test section of KNN algorithm
implementation and conclusion of the proposed system.
II.
SYSTEM OVERVIEW
In this section, we will describe an overview of the system
to be created that proposed in this research. Details of the
description of this system can be seen in Figure 1.
Government Departments using KNN Algorithm
Suhatati Tjandra*,
Amelia Alexandra Putri Warsito**
Department of Informatics Engineering
Sekolah Tinggi Teknik Surabaya
Surabaya, Indonesia
*[email protected], **[email protected]
Abstract—Participation of citizens in the process of city
development is very important. To achieve good governance and
democratic, the citizen can participate by providing complaints,
information, or advices. In the current system, complaints are
handled manually by 1-2 operators, whereas speed and accuracy
are needed. The problem is this manual handling causes errors in
the determination of appropriate government departments that
handle the complaint. This research will propose a system that
aims to determine the appropriate government department with
complaints given by the citizen with the implementation of KNearest Neighbor (KNN) algorithm, to reduce human errors.
This algorithm is one of text classification algorithms, which in
this research, is used to classify complaints which the texts in
Indonesian language. The input of the system is complaint given
by the citizen and the output is the name of the appropriate
government department, which is in accordance with the contents
of the complaint.
Keywords—text classification, k-nearest neighbor, supervised
learning, indonesian language, complaint
I.
INTRODUCTION
The growth of information technology that is rapidly
increasing makes the elements in it are affected. This
information technology greatly affects the processing,
distribution, keeping and communicating of information.
Existing institutions, such as companies, organizations, and
governments require information in the development of
institution itself, especially information that can support the
progress of an institution. However, to communicate this
information is sometimes not supported by good infrastructure,
so that there is no communication running smoothly.
Citizen likes to give complaints against what has happened
to the related institution. Citizen does not only convey a
complaint, but there also conveyed appreciation for what
happened. To accommodate complaints from the citizen, there
are institutions that provide the facilities to convey complaints
through various communication media. However, the
submitted complaints are sometimes not given a response and
there is a given response, but the response is given in a long
time. This happens because the institution is quite difficult to
determine the appropriate department in accordance with the
content of the complaint, because the relevant departments
should be determined manually and sometimes complaints
Judi Prajetno Sugiono
Department of Industrial Engineering
Sekolah Tinggi Teknik Surabaya
Surabaya, Indonesia
[email protected]
given to the departments which are not appropriate. This causes
responses given in a long time.
Based on the explanation, in this research, we propose a
system that can be used to handle these problems. A
classification algorithm in machine learning will be
implemented in this system. Arthur Samuel, who is a pioneer in
the field of machine learning, in 1959, defines that machine
learning as the field of study that gives computers the ability to
learn without being explicitly programmed [1]. One of the
categories of machine learning is a classification. In this
research, classification is conducted on the text, so it is called
as text classification or text categorization. The goal of the text
classification is to classify documents into a number of
predefined categories. There are many algorithms have been
developed for text classification, such as K-Nearest Neighbors
(KNN), Naïve Bayes, Decision Tree, Support Vector Machine
(SVM), etc [2].
Among all these algorithms, KNN is a widely used text
classifier because of its simplicity and efficiency [3]. Many
researches have shown that the KNN algorithm produces very
good performance in the experiments. In the research of short
text classification, KNN algorithm shows a better performance
compared with the Naïve Bayes and SVM algorithm [4].
Moreover, the other research, that uses Reuters corpus as a
dataset, shows that the KNN algorithm also produces very good
performance, when compared with the Naïve Bayes, Rocchio,
and C4.5 algorithm [5]. Based on that, we propose a system
that implements KNN algorithm to classify the text of the
complaint into a number of the government departments.
In this section, we explain the background and purpose of
this research. In the second section, we explain the proposed
system along with the input and output of the system. A
detailed explanation of KNN algorithm to determine the
appropriate government departments in accordance with the
contents of the complaint will be outlined in the third section.
The fourth and fifth sections are test section of KNN algorithm
implementation and conclusion of the proposed system.
II.
SYSTEM OVERVIEW
In this section, we will describe an overview of the system
to be created that proposed in this research. Details of the
description of this system can be seen in Figure 1.