Materials Today: Proceedings 64 (2022) 448–451

Contents lists available at ScienceDirect

Materials Today: Proceedings

journal homepage: www.elsevier.com/locate/matpr

COVID-19 vaccine sentiment analysis using public opinions on Twitter

P. Chinnasamy a,⇑, V. Suresh b, K. Ramprathap c, B. Jency A. Jebamani d, K. Srinivas Rao e, M. Shiva Kranthi f
a
Department of Computer Science and Engineering, MLR Institute of Technology, Hyderabad, India
b
Department of Computer Science and Engineering, Dr.N.G.P. Institute of Technology, Coimbatore, India
c
Department of Management Studies, M.Kumarasamy College of Engineering, Karur, India
d
Department of Computer Science and Engineering, KPR Institute of Engineering and Technology, Coimbatore, India
e
Department of Computer Science and Engineering, MLR Institute of Technology, Hyderabad, India
f
UG Student, Department of Computer Science and Engineering, MLR Institute of Technology, Hyderabad, India

a r t i c l e i n f o a b s t r a c t

Article history: Twitter, as is well known, is one of the most active social media platforms, with millions of tweets posted
Available online 28 April 2022 every day, in which different people express their opinions on topics such as travel, economic concerns,
political decisions, and so on. As a result, it is a useful source of knowledge. We offer Sentiment Analysis
Keywords: using Twitter Data for the research. Initially, our technology retrieves currently accessible tweets and
Public sentiments hashtags about various types of covid vaccinations posted on Twitter through using Twitter’s API.
Tweets Following that, the imported Tweets are automatically configured to generate a collection of untrained
Hashtags on covid vaccines
rules and random variables. To create our model, we’re utilizing, Tweepy, which is a wrapper for
Sentiment analysis
Machine learning algorithms
Twitter’s API. Following that, as part of the sentiment analysis of new Messages, the software produces
donut graphs.
Copyright Ó 2022 Elsevier Ltd. All rights reserved.
Selection and peer-review under responsibility of the scientific committee of the International Confer-
ence on Advanced Materials for Innovation and Sustainability.

1. Introduction used sentiment analysis to evaluate public transit customers and

got replies and feedback. Users’ responses and feedback are extre-
Passengers’ thoughts on Twitter can be examined using senti- mely varied, and each has their own point of view. NLTK is used for
ment analysis. In a study, sentiment analysis is used to investigate classification because it is an optimal method for classifying text,
people’s interests, assessments, and sentiments depends on what and it was chosen to ease the difficulty of the material on internet,
they wish to be written. Sentiment analysis is used as it has to deal which can be quite extensive. NLTK was also chosen based on the
with opinion, reality, and decisions that has to be taken. It is not results of a survey on algorithms that would be employed in this
uncommon for people to seek out the opinions of others when process. The output demonstrates that the NLTK is one of suitable
making decisions. Sentiment analysis is used in this study to look methods for solving sentiment analysis challenges. Furthermore,
at ideas of people who wrote on Twitter. The goal is to see how the decision tree technique was considered above the others since
commuter line passengers react to the Covid-19 being transmitted. it gave superior accuracy. The decision tree is also more optimal to
The twitter data was evaluated. Furthermore, the information utilize. Both methods are simple to use and produce accurate
gathered was divided into three categories by IEEE 2020. This arti- results. As a result, the nave Bayes method and the decision tree
cle is available for download and available, as well as complete text will be used in this study to discover which method produces
and data mining rights, as well as re-use and analysis rights for, superior accuracy. In this study, the writing system contains that
namely Positive, Negative, and Neutral. After that, sentiment anal- summarizes the study’s context and aims. The literature review
ysis can be used to determine people’s true influence. highlights the hypotheses that have been to aid in the develop-
The NLTK approach will be employed in this investigation. Sen- ment of this study. The research procedure explains how the data
timent Analysis of Public Transportation on social media Using the was gathered. Various methodologies can be used to perform sen-
NLTK was the method chosen based on past research. This study timent categorization. We can categorize these approaches in sev-
eral ways. in the three sorts that follow: 1) A lexicon-based
⇑ Corresponding author. strategy; 2) a machine learning/deep learning strategy, and 3) a
E-mail address: chinnasamyponnusamy@gmail.com (P. Chinnasamy). hybrid strategy.

https://doi.org/10.1016/j.matpr.2022.04.809
2214-7853/Copyright Ó 2022 Elsevier Ltd. All rights reserved.
Selection and peer-review under responsibility of the scientific committee of the International Conference on Advanced Materials for Innovation and Sustainability.
P. Chinnasamy, V. Suresh, K. Ramprathap et al. Materials Today: Proceedings 64 (2022) 448–451

2. Related works rithms, the MNB had the lowest accuracy of 88 percent. Marcec
and Likic [8] introduced vaccination drive analysis using the Twit-
Sentiment analysis is a technique for analyzing user’s thoughts, ter based sentiment analysis. Here they have used social network-
feelings, and judgments. According to some viewpoints, sentiment ing sites like Twitter might very well prove to be a useful medium
analysis is used to determine what users believe based on informa- of communication on SARS-CoV-2 vaccinations feelings and beliefs
tion such as written opinions. The analysis was carried out on the which can be observed nearly instantly.
written opinion, according to the two opinions. People frequently Alam et al [9], they have used valence aware dictionary for sen-
express and post their ideas on social media as a result of the dig- timent reasoned, a natural language processing (NLP) tool, was
ital era’s development, which makes us unable to avoid it. When it used to assess people’s feelings towards certain vaccines (VADER).
comes to internet, Twitter is one of the most popular places for They were able to illustrate the entire scenario by grouping the
people to express themselves. This can be utilized as a data source received attitudes into three categories (positive, negative, and
for analysis. Because of its prominence, Twitter was chosen as a neutral). We found that 33.96 percent of the respondents were
source of opinion mining. A concise explanation of beliefs and per- positive, 17.55 percent were negative, and 48.49 percent were neu-
ceptions up to 150 characters long on Twitter. This can be utilized tral. Nezhad et al [10,11], investigated Iranian people’s views on
as information in sentiment analysis to generate information and COVID-19 vaccinations by analyzing Iranian twitter posts. Discuss
determine people’s true influence. Text mining, which is a way of the various Iranian attitudes toward indigenous and imported
extracting important information from a material, was used in this COVID-19 vaccinations. They have used CNN-LSTM architecture,
sentiment analysis study. Text mining is that entails document we determined emotions of recovered tweets. Furthermore, we
preparation. Preprocessing transforms text from unstructured doc- looked into Iranian attitudes on the COVID-19 vaccine.
uments into structured information. The structured information is Samira et al [4], developed an analysis model based on the con-
subsequently classified using a data mining classification approach tent of Twitter. They were able to determine popular attitude and
[1]. thoughts about the COVID-19 vaccinations. They have discovered
The second study [2] used Naive Bayes, Decision Trees, and Ran- vaccination feelings and views in retweets after separation and
dom Forest to analyze sentiment in social media. Positive, negative, analyzed their growth over time, geographical location, principal
and neutral sentiments are categorized based on Twitter data. As a themes, hashtags, post analytics tools, as well as accounts features.
result, Indonesian Twitter users are more neutral in their views. Such tweets are collected from Twitter using a Twitter API authen-
The three methods with the highest accuracy were compared using tication token in this investigation. NLP is used to receive and ana-
data from fast miner tools. With an accuracy of 86.43 percent, Nave lyze the raw twitter posts. A supervised KNN clustering algorithm
Bayes is more accurate than the other method. The Decision Tree is then used to classify the produced data. The data is divided into
and Random Forest were both 82.91percent accurate. three categories by the algorithm: positive, negative, and neutral.
The third study [3] looked at classification with the Naive Bayes These categories pertain to the prevailing attitude of the persons
Classifier and Decision Tree Algorithms. The Naive Bayes classifica- whom Retweets are being analysed by Javed et al [12]. Ansari
tion and decision tree method are suggested for this study as a et al [13], they have analysed COVID-19 immunization retweets
more accurate and effective prediction in the assumption sort of to offer an assessment of the government’s sentiments to ongoing
brain tumor. The goal of this research is to show that the Decision immunization drives. Investigations on sentiment classification
Tree algorithm is more straightforward and straightforward than were also conducted in order to reveal new knowledge regarding
the Nave Bayes algorithm. The type of tumor has been identified the effects of location and gender including [3,14,15].
using these two algorithms, and it allows for the examination of As a result of the best approach that can be taken is that the
historical data from data sets, allowing neurologists to make edu- Naive Bayes algorithm is used for selection for training web pages.
cated guesses about the type of tumor. The Decision Tree algorithm
is, without a doubt, faster and more accurate than the Naive Bayes 3. Proposed method
algorithm. Furthermore, the fourth study uses Decision Tree, K-
Nearest Neighbor, and Nave Bayes to analyze sentiment from Panda’s library was used to turn the data-set into a data frame
tweets. A tweet from e-commerce is used as a case study in this for the purposes of this paper, making data manipulation and anal-
study. Text mining, text preprocessing, categorization, and other ysis easier. The Natural Language Toolkit (NLTK) is a computational
approaches are used in this study. These strategies are used to linguistics package that provides easy-to-use interfaces to over 50
develop sentiment analysis classification and analysis. Rapid miner lexical resources. This library was used for text pre-processing
is also utilized to aid in the comparison of analysis attitudes by uti- (stemming, lemmatization, and stop word removal), tokenization,
lizing three different categories inside the dataset. According to the sentiment analysis (NLTK Sentiment Intensity Analyzer), and Emo-
findings of this study [4,5], the Nave Bayes algorithm has the max- tional Analysis (Lexicon based) in this article. Matplotlib is a data
imum accuracy of 77 percent. The fifth study compares Naive visualization package that may be used to create static, animated,
Bayes, Decision Trees, and Neural Networks in classification. This and interactive data visualizations. This library was used to plot
study categorizes web pages through a of the three algorithms. time series graphs and donut chart for the purposes of this paper.
According to the findings of this study [6], they improved Nave. Tokenization is the process of dividing a large chunk of text into
The Bayes classifier not only out performs the traditional Nave smaller tokens. In this case, tokens can be words, characters, or sub
classifier. Bayes, on the other hand, performs similarly well. words. I utilized NLTK’s treebank tokenizer to perform word level
Khakharia et al [7], they have used Machine Learning to scan tokenization on tweets in this project. Text Pre-processing cleans
public tweets from Twitter connected to COVID-19 immunizations the text data and removes superfluous elements such as stop
in order to determine the user’s perspective on the vaccine (ML). words, hashtags, and emoticons, among other things, so that the
The analysis of the Twitter data into positive or negative messages text is ready to be given to the model.
was done using three distinct machine learning algorithms such as These methods were used in this study. Because of the strong
Multinomial Nave Bayes (MNB), Support Vector Machine (SVM), assumptions and great accuracy, the Naive Bayes method is used.
and Logistic Regression (LR). The LR algorithm produced the best The Decision Tree method was chosen because it is more straight-
results in the analysis, with a 97.3 percent accuracy, while SVM forward and straightforward to utilize. The following are the expla-
had a 96.26 percent accuracy. In comparison to the other ML algo- nations for the two methods:

449
P. Chinnasamy, V. Suresh, K. Ramprathap et al. Materials Today: Proceedings 64 (2022) 448–451

The proposed method starts with a data collection from differ-

ent sources like twitter, Gaggle, GitHub etc. After that, we are
applying tokenization methods on the dataset to identify the pos-
itive, negative and neutral values as shown in Figs. 1-3 respec-
tively. Fig. 4 explain about the proposed system flow of our
research. After tokenization, there are two processes going to hap-
pen like data preprocessing to remove the unwanted words from
the data’s, text processing is another task to identify the different
values in the input data. Based on this output the results should
passed into the cloud, then sentiment analysis is going to happen
by applying the decision tree algorithm.

Fig. 1. The positive value.

4. Performance evaluation

To summarize the findings, we can say that vaccines have had a

beneficial impact on the population, with the majority of individu-
als being either neutral or positive about the various vaccines
available. Individuals have a high level of trust in these vaccines,
but there are still angry and fearful feelings in the community
because many people are still infected with the deadly coronavirus,
and many of these immunizations have had bad side effects. The
majority of the tweets in this data set are from India, the United
States, the United Kingdom, and China, as these are the nations
that have been most afflicted by the virus or have been most
involved in the development of coronavirus vaccines.
After applying the decision tree algorithm to analyze the
covid’19 vaccines information. Here, we applied unigram and
Fig. 2. The negative value.

Fig. 3. The neutral value. Fig. 5. (a) (b) the unigram and Bigram of each sentiment.

Fig. 4. The architecture of the proposed method.

450
P. Chinnasamy, V. Suresh, K. Ramprathap et al. Materials Today: Proceedings 64 (2022) 448–451

demic, they might use sites such as twitter to spread good mes-
sages and reduce negative ones.
P.Chinnasamy, Writing the papers and responsible for
Implementation.
V.Suresh, Collection of DataSet.
K.Ramprathap - Proof Reading this article.
B Jency A Jebamani - Support for Plagiarism removal.
K. Srinivas Rao – Article writing.
Shiva Kranthi – Results and discussion section writing.

Declaration of Competing Interest

The authors declare that they have no known competing finan-

cial interests or personal relationships that could have appeared
to influence the work reported in this paper.

References

[1] A. De Figueiredo, C. Simas, E. Karafillakis, P. Paterson, H.J. Larson, Mapping

Fig. 6. The proposed method analysis using decision tree. global trends in vaccine confidence and investigating barriers to vaccine
uptake: a large-scale retrospective temporal modelling study, Lancet 396
(10255) (2020) 898–908, https://doi.org/10.1016/S0140-6736(20)31558-0.
[2] A. Hussain, A. Tahir, Z. Hussain, et al., Artificial intelligence-enabled analysis of
public attitudes on facebook and twitter toward COVID-19 vaccines in the
United Kingdom and the United States: observational Study, J. Med. Internet
Res. 2021;23(4):e26627. Published 2021 Apr 5. doi:10.2196/26627.
[3] K.S. Prasad, N.C.S. Reddy, B.N. Puneeth, A framework for diagnosing kidney
disease in diabetes patients using classification algorithms, SN Comput. Sci. 1
(2020) 101, https://doi.org/10.1007/s42979-020-0096-7.
[4] S. Yousefinaghani, R. Dara, S. Mubareka, A. Papadopoulos, S. Sharif, An analysis
of COVID-19 vaccine sentiments and opinions on Twitter, Int. J. Infect. Diseases
108 (2021) 256–262.
[5] Nhan Cach Dang, María N. Moreno-García, Fernando De la Prieta, Sentiment
analysis based on deep learning: a comparative study, Electronics 9(3) (2020)
483.
[6] Yong Shi et al., Survey on classic and latest textual sentiment analysis articles
and techniques, Int. J. Inform. Technol. Decis. Making 18(04) (2019) 1243-
1287.
[7] Aman Khakharia, Vruddhi Shah, Pragya Gupta, Sentiment analysis of COVID-19
vaccine tweets using machine learning (June 18, 2021). Available at http://
dx.doi.org/10.2139/ssrn.3869531.
[8] R. Marcec, R. Likic, Using Twitter for sentiment analysis towards AstraZeneca/
Oxford, Pfizer/BioNTech and Moderna COVID-19 vaccines Postgraduate
Medical, J. Publish. Online First 09 (2021), https://doi.org/
10.1136/postgradmedj-2021-140685.
[9] K.N. Alam, M.S. Khan, A.R. Dhruba, M.M. Khan, J.F. Al-Amri, M. Masud, M.
Rawashdeh, A. Korobeinikov, Deep learning-based sentiment analysis of
COVID-19 vaccination responses from Twitter data, Comput. Math. Methods
Med. 2021 (2021) 1–15.
Fig. 7. Different countries status of Covid’19 vaccines. [10] Z. Bokaee Nezhad, M.A. Deihimi, Twitter sentiment analysis from Iran about
COVID 19 vaccine, Diabetes Metab Syndr. 16 (1) (2021), https://doi.org/
10.1016/j.dsx.2021.102367 102367.
[11] Z. Bokaee Nezhad, M.A. Deihimi, Twitter sentiment analysis from Iran about
bigram techniques to apply for the each and every sentiment COVID 19 vaccine, Diabetes Metabol. Syndrome: Clin. Res. Rev. 16 (1) (2022)
parameter as shown in Fig. 5. 102367.
Fig. 6 shows the proposed method analysis of sentiment on [12] F.M. Javed Mehedi Shamrat, Sovon Chakraborty, M. M. Imran, Jannatun Naeem
Muna, Md. Masum Billah, Protiva Das, Md. Obaidur Rahman, Sentiment
twitter-based account using decision tree algorithm. This method analysis on twitter tweets about COVID-19 vaccines using NLP and supervised
provides high accuracy (96%) compare to other methods like linear KNN classification algorithm, Vol 23, No 1, pp. 463-470, 2021.
regression, classification etc. Fig. 7, clearly shows that the different [13] M.T.J. Ansari, N.A. Khan, Worldwide COVID-19 Vaccines sentiment analysis
through twitter content, Electron. J. Gen Med. (2021), https://doi.org/
countries of covid’19 vaccines details with sentiment analysis.
10.29333/ejgm/11316.
[14] K. Sai Prasad, O. Subhash Chander, G. Prabhakar Reddy, S. Gururaj, Artificial
5. Conclusion intelligence approach for classifying molecular dataset using density based
technique with appropriate Euclidean distance measure, Mater. Today:. Proc. 4
(8) (2017) 8827–8836.
In recent months, there has been a growth in unfavorable opin- [15] K.S. Prasad, S. Pasupathy, P. Chinnasamy, A. Kalaiarasi, An approach to detect
ion about both domestic and export vaccines in all over the globe, COVID-19 disease from CT scan images using CNN - VGG16 model,
International Conference on Computer Communication and Informatics
which is concerning. Because health authorities want to improve (ICCCI) 2022 (2022) 1–5, https://doi.org/10.1109/ICCCI54379.2022.9741050.
the adoption of COVID-19 immunizations in order to stop the epi-

451