COMMUNITY RESPONSE TO ELECTION POLICY DURING THE

COVID-19

Arya Arief Budiman*, Mujiono Sadikin

Department of Informatics, Faculty of Computer Science, Universitas Mercu Buana

Jl. Raya Meruya Selatan, Kembangan, Jakarta, Indonesia 11650

*Corresponding author, e-mail: [email protected],

[email protected]

Abstract – The purpose of this study is to I. INTRODUCTION

identify sentiment analysis of public responses
to the Indonesian Government's policy to carry The topic of the selected Final Project
out regional head elections (pilkada) during regarding Data Mining is classification in
the COVID-19 pandemic using the Naïve determining public responses in tweet data to
Bayes Classifier (NBC) and Support Vector government policy issues regarding the
Machine (SVM) algorithms. The research implementation of PILKADA (Regional Head
method used in this study is to use quantitative Elections) during the COVID-19 pandemic
research methods. The data used in this study using the Naïve Bayes Classifier (NBC)
were taken from public comments on a tweet in Algorithm Method, Support Vector Machine.
a twitter post that was saved in .csv format. Since the beginning of 2020 until now no one
The results obtained from this study are to can ignore the outbreak of Covid-19. It is not
compare 2 (two) algorithms, namely Naïve only countries that are preoccupied by this
Bayes and SVM into 3 test scenarios. The test sudden phenomenon, but also organizations,
results show the accuracy value obtained by even individuals. Initially, many parties were
SVM is much better than Naïve Bayes with the still in the denial stage, assuming that what
value in the first scenario Nave Bayes 76% : happened in Wuhan was only local, and
88% SVM, the second scenario Nave Bayes automatically disappeared as happened in
76% : 88% SVM, and the third scenario Nave previous experiences.
Bayes 78% : 90 % SVM. There are several
limitations in this study, such as the data used The Covid-19 phenomenon has triggered
only from the twitter platform, the data used various analyzes of the impact that occurred
only focuses on Indonesian posts, and only 2 and the response that needs to be prepared, so
(two) sentiment classification class labels are that the world is better prepared in the future to
used, namely positive and negative. This face similar shocks in the future. Various
research can be used as a reference by the efforts have been made to overcome this, from
General Election Commission (KPU) to domestic initiatives, to bilateral cooperation,
determine future regional head election but have not shown significant results.(Irawan,
policies during the covid-19 period. This 2020). Government is a form of organization
research falls into the area of data mining and dedicated to running a system of government.
machine learning disciplines. One of the government's tasks is to
disseminate information to the public through
Keywords: Election Policy, Naïve Bayes, Support conventional media, traditional media and new
Vector Machine, Algorithm, Sentiment media, to convey policies, work plans and
performance results (Menteri et al., 2013).

1
Dwi Adriansah, Head of National Industry training process (Arsya Monica Pravina, Imam
Twitter Indonesia, the average daily Twitter Cholissodin, 2019).
user worldwide in 2019 increased by 21%
year-on-year. Indonesia's growth ranks in the A. Problem Formulation
top five in the world(Putri, 2020). The 2020
Regional Head Election Policy (Pilkada) was Based on the background above, the writer
held during the COVID-19 Pandemic. There tries to formulate the problems that will be
are 270 districts that elect Regional Leaders. raised and discussed in this paper as follows:
The Pilkada phase starts from the end of
August to December 2020. 1. How to identify the sentiment analysis of the
public response to the Indonesian
The PILKADA simultaneously opens the door Government's Policy to carry out the
to the potential for greater spread of the PILKADA during the covid-19 pandemic?
COVID-19 virus in the community. Therefore,
it needs to be postponed until the pandemic 2. How is the application of the classification
subsides, and elections can be held method in this study using the Naive Bayes
simultaneously. However, on the other hand, Classifier (NBC) algorithm, and the Support
"PILKADA" must be carried out so that Vector Machine?
regional leaders can handle the pandemic most
effectively (Kompas, 2020). Judging from the 3. How is the performance comparison of the
election policies implemented by the classification method based on the accuracy of
government, during the COVID-19 pandemic the data using the Naive Bayes Classifier
the elections will be held simultaneously in (NBC) algorithm, and the Support Vector
2020, and the community has responded to the Machine?
policy.
B. Purpose and Benefits
Sentiment analysis or opinion mining is the
process of understanding, extracting and A. Purpose
processing textual data automatically to obtain
sentiment information contained in an opinion The research objectives are as follows:
sentence. Sentiment analysis is carried out to 1. To find out the identification of sentiment
see opinions or opinion tendencies towards a analysis of the public response to the
problem or object by someone, whether they Indonesian Government's policy to carry out
tend to have negative or positive views or PILKADA during the covid-19 pandemic.
opinions. One example of the use of sentiment
analysis in the real world is the identification 2. To find out the application of the
of market trends and market opinions on an classification method in terms of research
object of goods. The magnitude of the using the Naive Bayes Classifier (NBC)
influence and benefits of sentiment analysis algorithm, and the Support Vector Machine.
causes research and applications based on
sentiment analysis to develop rapidly. 3. To determine the performance of the
classification method from the results of the
Even in America there are about 20-30 level of accuracy to the data using the Naive
companies that focus on sentiment analysis Bayes Classifier (NBC) algorithm, and the
services (Liu, 2010). Naïve Bayes Support Vector Machine.
classification is a classification that is
supervised learning because it has a supervisor B. Benefits
(humans perform manual classification on the
data used in training) as a teacher in the The benefits of research are as follows:
learning process (Gunawan et al., 2018). 1. Can find out information about the response
Support Vector Machine (SVM) includes or public opinion in responding to the
machine learning (supervised learning) that Indonesian Government's policy to carry out
can predict classes based on the results of the the PILKADA during the covid-19 pandemic.

2
2. Can be used as a reference or input for the 4. Using 2 class labels for sentiment
government which can later be used to assess classification, namely positive and negative.
the effectiveness of government policies,
namely the Pilkada as an effort to continue the 5. The positive class label includes comment
election of a regional leadership during a data in the form of criticism and suggestions
pandemic emergency situation in the future. In that are supportive of the policy.
addition, the reference can be used as an
example of application for other countries in 6. The negative class label includes comments
an effort to carry out the upcoming that do not support the government and judge
Regional/General Election during the current that the policy has failed or is useless.
pandemic.
Using the classification method with the Naive
3. Can provide an overview of research to find Bayes Classifier (NBC) algorithm, and the
out whether there is a relationship between the Support Vector Machine.
government's policy efforts and current
community conditions such as the decline in
the positive number of Covid-19 per day in the D. THEORETICAL BASIS
community during the effort to carry out the On the basis of theory can
PILKADA. explain related research, then
supporting theories related to
4. Can provide an overview of research in
sentiment analysis of community responses to
research.
the case study using the Naive Bayes Classifier
(NBC) algorithm, and the Support Vector A. Related Research
Machine.
Related research can be
5. Can provide an overview of research in the explained using tables and explained
comparison of classification performance from briefly in the form of paragraphs or in
the results of sentiment analysis of community the form of paragraphs as a whole as a
responses to the case study using the Naive
whole completely. The procedure for
Bayes Classifier (NBC) algorithm, and the
Support Vector Machine. writing table labels can be seen in Table
1.
C. Problem Limitation

Limitation of research problems in terms of

writing and research as follows:
1. The data used regarding the 2020 Pilkada /
Pilkada during the covid-19 pandemic were
taken from social media Twitter.

2. The data used as research focuses on

Indonesian-language tweets on government
policy issues regarding carrying out regional
elections during the Covid-19 pandemic.

3. The data retrieval process (crawling data)

lasts for approximately 5 months which
includes election data from September 23 to
January 23.

3
Topik Metode Hasil Referensi
Analisis Lexicon The average accuracy results obtained for the (Buntoro,
Sentimen Based, Naïve Bayes method are 74.6% and 75.62% 2017)
Calon Support for the SVM method. The best results were
Gubernur DKI Vector obtained similar to those obtained when using
Jakarta 2017 Machine, the experiment in English. The results of the
Di Twitter Naïve Bayes classification process are then tested using the
10 fold cross validation method, the data is
divided into 10 parts with 9/10 parts used for
the training process and 1/10 parts used for the
testing process.
Analisis Support Vector In the sentiment analysis research on (Rofiqoh
Sentimen Tingkat Machine dan the level of satisfaction of users of et al.,
Kepuasan Lexicon Based cellular telecommunications service 2017)
Pengguna Features providers on Twitter with the Support
Penyedia Vector Machine and Lexicon Based
Layanan Features methods, the accuracy value
Telekomunikasi is 79%, precision is 65%, recall is
Seluler Indonesia 97%, and f-measure is 78%. This
Pada Twitter proves that sentiment analysis using
Dengan Metode the Support Vector Machine and
Support Vector Lexicon Based Features methods is
Machine dan good for classifying sentiment
Lexicon Based analysis.
Features

Analisis Sentimen K-Nearest The textual weighting (Nurjanah et

Terhadap Tayangan Neighbor dan results from the K-Nearest al., 2017)
Televisi Berdasarkan Pembobotan Neighbor classification and
Opini Jumlah Retweet the non-textual weighting
Masyarakat pada Media from the weighting of the
Sosial Twitter number of retweets will be
menggunakan Metode combined using certain
K-Nearest Neighbor dan constant values (α and ) to
Pembobotan Jumlah produce positive and
Retweet negative sentiments. The
data used in the form of
public opinion on television
shows on twitter amounting
to 400. From the results of
accuracy testing using
textual weighting obtained
82.50%, using non-textual
weighting 60%, and using a
combination of both
83.33% with a value of k =
3 and constants The correct
multiplication is =0.8 and

1
 =0.2. word.

Machine Learning-Based Methods of the results with the different (Hasan et al.,
Sentiment Analysis for sentiment analysis classifiers. The experiment 2018)
Twitter Accounts can be categorized on Twitter data will show
predominantly, as which technique has a better
machine-learning, capability of measuring
Lexicon-based, and sentiment prediction
hybrid. accuracy. The remainder of
the paper is organized in the
following sections.
Additionally, we tested their
results with two supervised
machine-learning
classifiers, Naïve Bayes and
SVM. Although the results
of TextBlob were relatively
better, we obtained the best
result when analyzing
tweets with W-WSD, as is
clearly shown in the results.
In order to take our
initiative to the next level,
we will find the patterns of
political parties based on
Twitter reviews in future
research.

2
Analisis Sentimen Metode Naïve The very rapid growth of (Prananda
Tentang Opini Film Pada Bayes social media does not make Antinasari,
Dokumen Twitter Twitter abandoned by its Rizal Setya
Berbahasa users. Twitter is one of the Perdana,
Indonesia Menggunakan social media that allows 2017)
Naive Bayes Dengan users to interact, share
Perbaikan Kata Tidak information, or even to
Baku express feelings and
opinions, including
expressing opinions on
films. Comments or Tweets
about films on Twitter can
be used as an evaluation in
watching films and
improving film production.
To find out, sentiment
analysis can be used to
classify into negative or
positive sentiment. Tweets
contain a variety of
languages used, including
non-standard forms of
language such as slang,
word abbreviations, and
misspellings. Therefore,
special handling is needed
on Tweets. In this study,
non-standard word
dictionary and Levenshtein
Distance normalization
were used to correct non-
standard words into
standard words with Naive
Bayes classification. Based
on the results of the tests
that have been carried out,
the highest accuracy was
obtained with the values of
accuracy, precision, recall,
and f-measure of 98.33%,
96.77%, 100%, and
98.36%.

3
Analisis Sentimen Metode K-Nearest The 2013 curriculum has (Mentari et
Kurikulum 2013 Pada Neighbor dan become a topic that is often al., 2018)
Sosial Media Twitter Feature Selection discussed by the public, one
Menggunakan Metode Query Expansion of which is on Twitter.
K-Nearest Neighbor dan Ranking Twitter is one of the most
Feature Selection Query popular social media and is
Expansion Ranking used by the public to
express opinions or
opinions about certain
topics. Thus, this study tries
to analyze tweets about the
2013 Curriculum by
classifying whether it is a
positive opinion or a
negative opinion. The
classification process is
carried out using the K-
Nearest Neighbor method
and using the Query
Expansion Ranking method
for feature selection. There
are 4 main processes in this
sentiment analysis system,
the first is text
preprocessing, term
weighting (TF-IDF), feature
selection, and classification.
Based on the tests that have
been carried out in this
study, it is proven that
feature selection increases
the accuracy of the system.
The best accuracy result of
96.36% is obtained when
the value of k = 1 and uses
a 50% feature selection
ratio. The test results using
a 50% ratio of feature
selection have a higher
accuracy than systems that
do not use feature selection
because some noise features
are removed.

4
Analisis Sentimen Metode Support With the increasing use of (Arsya
Tentang Opini Maskapai Vector Machine Twitter, social media that Monica
Penerbangan pada (SVM) works in real-time for the Pravina,
Dokumen Twitter public to express Imam
Menggunakan Algoritme complaints and appreciation Cholissodin,
Support Vector Machine for airlines, it is necessary 2019)
(SVM) to create a system that can
classify tweets containing
opinions which belong to
what class, in this study
there are positive and
negative classes. This is
done in order to assist
airline companies in terms
of evaluating service
improvements and to assist
the public in choosing the
right airline. So that
sentiment classification is
carried out with the Lexicon
Based feature which can
accept opinions in
languages other than
Indonesian (in this study
English is used) to conduct
sentiment analysis. The
support vector machine
algorithm is used to perform
classification. The results of
this study indicate the
optimal parameters and the
effect of using Lexicon
Based Features. By using
parameter C with a value of
10 and a learning rate of
0.03 and using Lexicon
Based Features with literacy
50 times, it gives 40%
accuracy, 40% precision,
100% recall, and 57.14% f-
measure.
ANALISIS SENTIMEN Metode Naïve Nowadays people cannot be (Rustiana &
PASAR OTOMOTIF Bayes separated from the internet, Rahayu,
MOBIL: TWEET from waking up in the 2017)
TWITTER morning most people are
MENGGUNAKAN already busy to update
NAÏVE BAYES status, browse news and
send email. In social

5
 networks such as Facebook
or others, most people tell
the atmosphere they
experience to social
networking sites such as
telling about being on
vacation, eating, traveling
on Facebook, Twitter or
others, so that from this data
it is now forming a very
large data or known as "Big
Data". Social media such as
Twitter is widely used to
pour out the hearts of its
users, so that the data can
describe sentiments. Naïve
Bayes is a sentiment
analysis classification
method that is widely used
because of its robustness,
simplicity and high
accuracy in classifying data.
In this study, we analyze
data from Twitter by
searching for the best-
selling car brands based on
data from GAIKINDO
(Indonesian Automotive
Industry Association). By
using three classes, it can
produce an accuracy value
of 93%, and a positive
precision level of 90%, a
negative precision of 90%
and a neutral sentiment of
100% so that nave Bayes is
able to analyze with high
accuracy.
supports what innovation is added
to the selected topic.
2. 1. Submission of support is carried
B. Teori Pendukung out by both the success team,
busers and supporters to have a
1. The theory that supports this positive image of each candidate.
research is about the innovation Various media are used, one of
process, so it is necessary to which is Twitter, people submit
elaborate on the theory that positive and negative comments
and even tend to "black campaign"

6
 and hoaxes before the election is input space (Rustiana & Rahayu,
held or when the election is in 2017).
progress regarding the election 8. The Naive Bayes classifier is a
being held, comments on Twitter at concept. Determine the probability
this time cannot be determined of a group of document categories.
more positively or negatively, Classification algorithms can
Therefore, it is necessary to handle large amounts of data and
conduct a sentiment analysis to have high precision results
determine the tendency of public (Rustiana & Rahayu, 2017).
opinion towards the election 9. As test data, the number of data in
(Kurniawan & Susanto, 2019). the data file is cut from the last
3. Classification is a process to find a record for each positive opinion
model from a data. The purpose of and negative opinion data. The
the classification process is to make training data was obtained from
a decision by predicting a case combining the positive opinion data
based on the results of the files and the remaining negative
classification that has been opinion data files from the use for
obtained (Kurniawan & Susanto, test data (Rustiana & Rahayu,
2019). 2017).
4. Currently, social networking sites
such as twitter have become a very
popular communication tool among II. METHODS
internet users. Millions of tweets
appear every day on the social In this study, quantitative research methods are
networking site Twitter. Twitter is used, where this type of research raises or
one of the most widely used social describes phenomena or research facts in
networks. Almost all people in digital form, or collects in-depth data and fully
various countries use it (Andrianto explains phenomena that occur in society. This
research also adopts an exploratory form,
& Adinugroho, 2018).
namely data obtained from research or field
5. Naïve Bayes Classifier is the exploration to obtain further information,
concept of probability of which includes behavior, perception,
determining document class motivation and behavior described in words
groups. This classification and language in a natural context. In this study,
algorithm can process large the completeness and depth of the data studied
amounts of data with high accuracy is very important. This study collects data by
results (Rossi et al., 2017). collecting public comments in the form of
6. There are 2 machine learning tweets on Twitter posts, which can be saved in
techniques, namely supervised .csv format.
learning and unsupervised learning.
The research flow is illustrated in Figure 1.
The difference between these two
Research Phase.
techniques is the presence or
absence of labels on the training
data (Rahmawati et al., 2017).
7. SVM is a machine learning method
based on the principle of structural
risk minimization (SRM) aiming to
find the best hyperplane that
separates two categories in the

2
 In the preprocessing stage, the following 4
steps are carried out.
 Data Cleansing cleans up hashtags,
usernames, URLs, retweets, numbers and
emoticons in tweets.

 Case folding converts uppercase letters to

lowercase letters for processing.

 Tokenizing splits the text and checks each

individual word.

 Stopwords removal removal of

conjunctions.

 Stemming forms the basic words after the

tokenizing process.

 Remove Duplicate is used to delete the

same data.

Figure 1. Research Methodology

A. Data collection

The data collection method used for the

research is to retrieve tweet data from Twitter, Figure 2. Before Preprocessing
this can be done by integrating the Twitter API
with Rapid Miner, data collection begins
before the election and after the election, the
2020 election is held on December 9, 2020,
Rapid Miner is based on query terms objects in
data mining processing applications that are
connected to the Twitter API. from the chirp
obtained contains "the government's efforts to
implement the election policy". For tweets
about PILKADA, the query data set includes
tweets from (#) and regular tweets. This tweet
is split or converted into several data sets, such
as the # Simultaneous Disaster Pilkada dataset,
2020 Pilkada (due to covid-19) and regional
head election policies during covid - 19.

Name Data Data Tweets

Pilkada 39077
Figure 3. After Preprocessing
Table 1. Twitter Data

B. Data Pre-Processing
C. Data Class Labeling

3
The next stage of data collection is data E. TF-IDF Weighting
labeling. Where at this labeling stage will be
done manually. For the results of the labeling Word weighting is the process of assigning a
in the form of positive sentiment and negative weight to each word contained in a document.
sentiment. Then do the calculations using the In searching for ranking information based on
Support Naïve Bayes Classifier (NBC) word frequency, one of the most popular
algorithm, and the Support Vector Machine. methods is the TF-IDF (Term Frequency -
First, namely Coarse-grained sentiment Inversed Document Frequency) method (Liu,
analysis, we try to carry out the analysis 2010).
process at the Document level. In short, we are
trying to classify the orientation of a document F. Classification
as a whole. There are 3 types of orientation:
Positive, Neutral, Negative. However, there are After data processing, labeling, and
also those who make this orientation value resampling, the next stage is the classification
continuous / not discrete. The second is the of sentiment analysis. This stage is the stage
Fine-grained sentiment analysis of this second for providing training and implementing
category which is currently on the rise. The various machine learning algorithms. Is the
point is that most of the researchers focus on stage of selecting mining techniques by
this type. The object to be classified is not at determining the algorithm to be used. This
the document level but a sentence in a study uses tools used to perform modeling in
document. accordance with predetermined techniques, the
example: tools are Python. The sentiment classification
I hate wet wipes. (negative) (-1) stage is the process of testing tweet data
My campus is very beautiful. (positive) (+1) classification using the classification algorithm
(Azkia, 2019). used, namely, the algorithm as the model. The
result of testing each model is to classify
Positif Negatif tweets.

Pilkada 8308 3990 Naïve Bayes is a classification method that is

often used in sentiment analysis because it is
Table 2. Classification Value simple and easy to classify documents. positive
and negative tweets to get the best accuracy
value for each algorithm (Kurniawan &
D. Resampling
Susanto, 2019). The concept of classification
with the Support Vector Machine is to find the
Applying the selected resampling model so
best hyperplane that functions as a separator of
that here there is a transformation of an
two data classes. Support Vector Machine is
unbalanced dataset into a balanced dataset for
able to work on high-dimensional datasets by
re-evaluation but only with 3 models with high
using the kernel trick (Rofiqoh et al., 2017).
accuracy (Wijanarto & Brilianti, 2020). Since
the sentiment in election tweets was
unbalanced, a random method was used for
resampling, which was useful for balancing the III. RESULTS AND DISCUSSION
data. The method used is to reduce most of the
data, namely downsampling. The following are In this study, the authors conducted
the classification values after downsampling: experiments on case studies of the 2020
regional elections during this pandemic using 2
Positif Negatif algorithms, Support Vector Machine and
Naïve Bayes. There are 3 scenarios that divide
Pilkada 3990 3990 the percentage of data. The distribution of data
presentation is the separation of training and
Table 3. Resampling Value testing data based on the percentage, for
example 90%: 10% means 90% is training data

4
and 10% is testing data. In this study, the following are the results of the distribution of
percentage distribution was divided into 3 the 80% performance percentage: 20%
experimental scenarios, namely the first
experiment using data sharing of 90%: 10%,
the second experiment using data sharing of
80%: 20% and the third experiment using data
sharing of 70%: 30%. Accuracy in each case
can be different because the concept of
confusion matrix will be used to evaluate the
model that has been formed.
The formula used to determine accuracy is:
Figure 6. Naïve Bayes 80:20
Accuracy = (TP + TN) / (TP + FP + TN + FN)
* 100
TP = True Positive
TN = True Negative
FP = False Positive
FN = False Negative

The results of the model performance using

data sharing or percentage split 90% : 10% For Figure 7. SVM 80:20
the case study the model results are, for the
SVM algorithm the accuracy is 88% for the
Naïve Bayes algorithm the accuracy is 76%. The results of the model performance using
Following are the results of the distribution of data sharing or percentage split 70% : 30% For
90% performance percentage: 10% the case study the model results are, for the
SVM algorithm the accuracy is 90% for the
Naïve Bayes algorithm the accuracy is 78%.
The following are the results of the distribution
of the 70% performance percentage: 30%

Figure 4. Naïve Bayes 90:10

Figure 8. Naïve Bayes 70:30

Figure 5. SVM 90:10

The results of the model performance using

data sharing or percentage split 80% : 20% For
case studies the model results are, for the SVM Figure 9. SVM 70:30
algorithm the accuracy is 88% for the Naïve
Bayes algorithm the accuracy is 76%. The

5
 IV. CONCLUSION Information Technology, 1(1), 32–41.
https://www.researchgate.net/profile/Ghu
The results of this study are experiments using lam_Buntoro/publication/316617194_An
the support vector machine algorithm have the alisis_Sentimen_Calon_Gubernur_DKI_J
highest accuracy value among the three akarta_2017_Di_Twitter/links/5907eee44
experimental scenarios, and a percentage split 585152d2e9ff992/Analisis-Sentimen-
is carried out for 1 data to be tested. The Calon-Gubernur-DKI-Jakarta-2017-Di-
results of the discussion are summarized in the Twitter.pdf
table. Gunawan, B., Pratiwi, H. S., & Pratama, E. E.
(2018). Sistem Analisis Sentimen pada
Pilkada Ulasan Produk Menggunakan Metode
Percobaan Naive Bayes. Jurnal Edukasi Dan
Naïve Bayes SVM
90:10 76% 88% Penelitian Informatika (JEPIN), 4(2),
80:20 76% 88% 113.
70:30 78% 90% https://doi.org/10.26418/jp.v4i2.27526
Table 4. Comparison of Accuracy Hasan, A., Moin, S., Karim, A., &
Shamshirband, S. (2018). Machine
From the following table, it can be concluded Learning-Based Sentiment Analysis for
that the 70:30 percentage split model has the Twitter Accounts. Mathematical and
best accuracy value on the test data, so it can Computational Applications, 23(1), 11.
be concluded that the 70:30 percentage split https://doi.org/10.3390/mca23010011
model using the support vector machine Irawan, A. J. H. (2020). Fenomena Covid-19:
algorithm is the best experimental plan. In this Dampak Globalisasi Dan Revitalisasi
research. Multilateralisme. Jurnal Ilmiah
THANK-YOU NOTE Hubungan Internasional, 0(0), 47–52.
The author would like to thank Mr. Mujiono https://doi.org/10.26593/jihi.v0i0.3877.4
Sadikin as the supervisor who has guided me 7-52
in this research. Kompas. (2020). Alasan Pro dan Kontra
Pilkada Serentak di Tengah Pandemi
REFERENCES Covid-19 Halaman all - Kompas.com.
Andrianto, B., & Adinugroho, S. (2018). https://www.kompas.com/tren/read/2020/
Analisis Sentimen Konten Radikal 09/24/072900565/alasan-pro-dan-kontra-
Melalui Dokumen Twitter Menggunakan pilkada-serentak-di-tengah-pandemi-
Metode Backpropagation. Jurnal covid-19?page=all
Pengembangan Teknologi Informasi Dan Kurniawan, I., & Susanto, A. (2019).
Ilmu Komputer, 2(12), 7380–7385. Implementasi Metode K-Means dan
Arsya Monica Pravina, Imam Cholissodin, P. Naïve Bayes Classifier untuk Analisis
P. A. (2019). Analisis Sentimen Tentang Sentimen Pemilihan Presiden (Pilpres)
Opini Maskapai Penerbangan pada 2019. Eksplora Informatika, 9(1), 1–10.
Dokumen Twitter Menggunakan https://doi.org/10.30864/eksplora.v9i1.23
Algoritme Support Vector Machine 7
(SVM). Jurnal Pengembangan Teknologi Liu, B. (2010). Sentiment analysis and
Informasi Dan Ilmu Komputer, 3(3), subjectivity. Handbook of Natural
2789–2797. http://j-ptiik.ub.ac.id Language Processing, Second Edition,
Azkia, M. W. Al. (2019). Sentiment Analysis 627–666.
In R. Assalamualaikum kawan.. | by Mentari, N. D., Fauzi, M. A., & Muflikhah, L.
Muhammad Wildan Al Azkia | Medium. (2018). Analisis Sentimen Kurikulum
https://medium.com/@wildanalazkia/sent 2013 Pada Sosial Media Twitter
iment-analysis-in-r-8924856e4b62 Menggunakan Metode K-Nearest
Buntoro, G. A. (2017). Analisis Sentimen Neighbor dan Feature Selection Query
Calon Gubernur DKI Jakarta 2017 Di Expansion Ranking. Jurnal
Twitter. INTEGER: Journal of Pengembangan Teknologi Informasi Dan
Ilmu Komputer (J-PTIIK) Universitas

6
 Brawijaya, 2(8), 2739–2743. Machine. CCIT Journal, 10(2), 197–206.
Menteri, P., Negara, P. A., Birokrasi, D. A. N. https://doi.org/10.33050/ccit.v10i2.539
R., Indonesia, R., Media, P. P., & Rofiqoh, U., Perdana, R. S., & Fauzi, M. A.
Pemerintah, S. I. (2013). 2013, No.102 4. (2017). Analisis Sentimen Tingkat
102, 4–24. Kepuasan Pengguna Penyedia Layanan
Nurjanah, W. E., Perdana, R. S., & Fauzi, M. Telekomunikasi Seluler Indonesia Pada
A. (2017). Analisis Sentimen Terhadap Twitter Dengan Metode Support Vector
Tayangan Televisi Berdasarkan Opini Machine dan Lexion Based Feature.
Masyarakat pada Media Sosial Twitter Jurnal Pengembangan Teknologi
menggunakan Metode K-Nearest Informasi Dan Ilmu Komputer (J-PTIIK)
Neighbor dan Pembobotan Jumlah Universitas Brawijaya, 1(12), 1725–
Retweet. Jurnal Pengembangan 1732. http://j-ptiik.ub.ac.id/index.php/j-
Teknologi Informasi Dan Ilmu Komputer ptiik/article/view/628
(J-PTIIK) Universitas Brawijaya, 1(12), Rossi, A., Lestari, T., Setya Perdana, R., &
1750–1757. Fauzi, M. A. (2017). Analisis Sentimen
Prananda Antinasari, Rizal Setya Perdana, M. Tentang Opini Pilkada DKI 2017 Pada
A. F. (2017). Analisis Sentimen Tentang Dokumen Twitter Berbahasa Indonesia
Opini Film Pada Dokumen Twitter Menggunakan Näive Bayes dan
Berbahasa Indonesia Menggunakan Pembobotan Emoji. Jurnal
Naive Bayes Dengan Perbaikan Kata Pengembangan Teknologi Informasi Dan
Tidak Baku. Jurnal Pengembangan Ilmu Komputer, 1(12), 1718–1724.
Teknologi Informasi Dan Ilmu Komputer, http://j-ptiik.ub.ac.id
1(12), 1733–1741. http://j-ptiik.ub.ac.id Rustiana, D., & Rahayu, N. (2017). Analisis
Putri, V. M. (2020). Tahun 2019, Twitter sentimen pasar otomotif mobil: Jurnal
Kembali Jaya di Indonesia. Detik Inet. SIMETRIS, 8(1), 113–120.
https://inet.detik.com/mobile-apps/d- Wijanarto, W., & Brilianti, S. P. (2020).
4905885/tahun-2019-twitter-kembali- Peningkatan Performa Analisis Sentimen
jaya-di-indonesia Dengan Resampling dan Hyperparameter
Rahmawati, A., Marjuni, A., & Zeniarja, J. pada Ulasan Aplikasi BNI Mobile. Jurnal
(2017). Analisis Sentimen Publik Pada Eksplora Informatika, 9(2), 140–153.
Media Sosial Twitter Terhadap https://doi.org/10.30864/eksplora.v9i2.33
Pelaksanaan Pilkada Serentak 3
Menggunakan Algoritma Support Vector