Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research

Technology
ISSN No:-2456-2165
https://doi.org/10.38124/ijisrt/IJISRT24APR227

Sentiment Analysis on Twitter through

Machine Learning: A Comprehensive
Approach with User-Centric Visualisations
Archis Khuspe1; Tejas Gaikwad2; Agnibha Sarkar3; Medha Wyawahare4; Ankita Kumari5; Abhay Chopde6
Department of Electronics and Telecommunication
Vishwakarma Institute of Technology Pune, India

Abstract:- Sentiment analysis is a crucial field that deals positive, negative, or neutral. This analysis is often carried
with the intricate task of identifying and systematically out using programming languages like R and Python,
categorizing the various perspectives and opinions making it a versatile tool for understanding sentiment on
expressed within the original text. In today's digital age, platforms like Twitter.
social media platforms serve as a prolific source of data, A. Problem Statement
inundated with a relentless stream of status updates, In today's digital age, social media platforms,
tweets, and content imbued with sentiments. Analysing particularly Twitter, serve as prolific sources of unceasing,
the sentiments conveyed by users in this vast reservoir of unfiltered, and real-time data. Among the myriad
data holds a pivotal role in comprehending the collective applications of this data, sentiment analysis stands out as a
sentiments of the user community, dissecting dialogues, pivotal tool for understanding and categorizing the vast
and aggregating viewpoints. This, in turn, can be spectrum of emotions and opinions expressed within this
instrumental in shaping strategies for commerce, digital ecosystem. While sentiment analysis on social media
conducting insightful political research, and gauging the data is widely recognized and employed in various domains,
pulse of communal activities. Examining sentiments on there exist several critical research gaps that warrant
Twitter presents an increased difficulty because of the attention. This research aims to address these gaps and
frequency of spelling errors, casual language, icons, and contribute to the field of sentiment analysis by focusing on
emojis. This research focuses on Twitter sentiment the following key challenges:
analysis, with a specific emphasis on a particular user
account. The approach involves a combination of Python  Handling Noisy and Informal Text
programming and Machine Learning techniques. By Social media platforms, especially Twitter, are rife
embarking on a comprehensive sentiment analysis with colloquial language, abbreviations, misspellings, and
journey within a specific domain, the aim is to discern emoticons, which pose challenges for accurate sentiment
the profound impact of that domain's data on sentiment analysis. This paper aims to develop effective techniques for
categorization. Furthermore, this paper introduces a preprocessing and cleaning such noisy data to improve the
novel feature that enhances the organization of a user's accuracy of sentiment categorization.
most recent tweets and their presentation through visual
aids such as graphs, charts, and word clouds. This  User-Specific Sentiment Analysis
visualization approach empowers a more intuitive and Existing sentiment analysis approaches often provide a
insightful exploration of the sentiments and trends broad overview of sentiment within a dataset. However, a
embedded within the user's Twitter activity, facilitating research gap exists in understanding how the sentiment of
a deeper understanding of their thoughts and emotions individual Twitter users evolves over time. This paper seeks
as expressed through their digital interactions. to address this gap by exploring user-specific sentiment
analysis, providing insights into the changing emotional
Keywords:- Twitter, Sentiment Analysis, Dataset, User landscape of individual users.
Accounts, Machine Learning Algorithms, Tweets, Data
Visualisation  Data Visualization for Enhanced Interpretation
While sentiment analysis results are valuable,
I. INTRODUCTION conveying these insights effectively to users is a challenge.
This paper aims to fill a research gap by introducing and
Sentiment analysis, a widely adopted method for evaluating various data visualization techniques, such as
extracting insights from text, plays a pivotal role in diverse word clouds, bar graphs, and pie charts, to enhance the
applications. It leverages innovative text mining techniques interpretability and usability of sentiment analysis results.
to assess the emotional tone of content, categorizing it as

IJISRT24APR227 www.ijisrt.com
2066
Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research
Technology
ISSN No:-2456-2165
https://doi.org/10.38124/ijisrt/IJISRT24APR227

 User Engagement and Sentiment Relationship by comparing them with various machine learning
Understanding the connection between sentiment and algorithms, specifically examining sentiment analysis in the
user engagement on Twitter is vital for businesses and distinct context of Twitter data. [4] Sentiment analysis is
individuals aiming to leverage the platform effectively. This employed to evaluate the model's efficiency in social media,
paper seeks to address the research gap related to the using standard performance metrics such as accuracy, recall,
relationship between sentiment and user engagement metrics precision, and F1 score to assess its effectiveness. [5]
like likes, retweets, and replies. Twitter sentiment analysis is utilized as an example to
expedite rapid decision-making in the FTSE stock market by
leveraging predictive abilities based on sentiment analysis
 Real-Time Sentiment Analysis results. [6] DICET (Data Integration and Correlation Engine
In the era of instant communication, real-time for Twitter) is employed to acquire and align data features to
sentiment analysis is essential for businesses and extract essential information, enhancing the quality of data
organizations. The paper aims to contribute to the research used in sentiment analysis. [7] A groundbreaking method for
gap surrounding the implementation and application of real- Qualitative Spatial Reasoning transforms semantic data
time sentiment analysis on social media platforms, with a analysis for private documents, highlighting the power of
focus on Twitter. using spatial relationships to understand confidential
content. [8] A cutting-edge sentiment analysis framework is
The approach utilized for sentiment analysis involves crafted through the implementation of advanced ML
an examination of the libraries and tools employed, the data techniques, notably Entropy maximization and Simple
retrieval process involving the Twitter API, and the data Bayes. This system excels in categorizing and scrutinizing
cleaning processes indispensable for the interpretation of sentiments conveyed in textual data. [9] Twitter data related
sentiment. This combines subjectivity and polarity metrics to the COVID-19 pandemic is analysed to reveal emotional
to assess the opinions and sentiments within the collection and attitudinal dimensions in public conversations during
of tweets, categorizing them as positive, negative, or neutral. this global health crisis. [10] A framework was applied to
The integration of visual representations, such as word recognize and categorize sentiments expressed by Twitter
clouds, bar graphs, and pie charts, augments the users related to a specific product or commodity, valuable
comprehensibility of the sentiment analysis outcomes, for understanding customer opinions and preferences. [11]
thereby presenting an all-encompassing depiction of the Using various machine learning methods, sentiment
emotional dynamics underpinning the user's Twitter activity. dispersion on Twitter is investigated, utilizing SentiDiff—a
novel method considering emotional reversals—to
The outcomes of the sentiment analysis provide a categorize sentiments, alongside developing a predictive
valuable insight into the user's emotional involvement with model. [12] The paper investigates how individuals express
Twitter. By discerning the emotional pulse of the user's feelings and moods amid the COVID-19 pandemic,
tweets, this analysis not only reveals the prevalent sentiment specifically on platforms like Twitter, to understand the
within the Twitter activity but also explicates the user's dynamics of their thoughts and feelings. [13] A novel
tendencies, preferences, and interactions with their digital method is introduced to enhance aspect reduction in
audience. This understanding is instrumental in a myriad of sentiment analysis, utilizing advanced mathematical analysis
applications, from enhancing marketing strategies and and n-grams to create a tailored Twitter sentiment lexicon.
public safety measures to predicting political outcomes and The approach leads to a substantial boost in sentiment
fostering a more empathetic and informed societal approach. classification accuracy. [14] Twitter sentiment analysis of
information concerning ordinal reversion is explained, likely
II. LITERATURE REVIEW focusing on understanding how sentiment changes over
time. [15] This research explores sentiment analysis
The security of the Kirchhoff-law-Johnson-noise’s automation for online opinions using machine learning
protected vital exchange system is assessed, emphasizing approaches (SVM, Naive Bayes, Decision Tree, KNN) and
the identification of vulnerabilities and strengthening the deep neural network (RNN-LSTM). RNN-LSTM achieves
overall security infrastructure. [1] New clustering algorithms the highest accuracy (88-93%) on Twitter datasets (IMDB,
are introduced, specifically K-means and DENCLUE, to Amazon, Airline) compared to other algorithms. [16] The
analyse sentiments expressed in tweets. These methods aim research paper introduces a unified framework for Twitter
to group similar sentiments for more in-depth analysis. [2] sentiment analysis, addressing classification challenges. The
Researchers scrutinize various pre-processing methods and proposed hybrid scheme incorporates slang, emoticon,
test them on two different datasets to evaluate their SentiWordNet, and domain-specific classifiers, improving
effectiveness in improving sentiment analysis results. [3] An accuracy by considering various linguistic elements in
advanced Recurrent Neural Network - Long Short-Term tweets. [17] The study employs the VADER Sentiment
Memory is deployed on the extensive Twitter database to Analyzer for multi-classification sentiment analysis of
categorize individuals' viewpoints as positive or negative Twitter data related to the 2016 US election, yielding
sentiments. The study evaluates the accuracy of these results accurate results. Limitations include small data volume and

IJISRT24APR227 www.ijisrt.com
2067
Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research
Technology
ISSN No:-2456-2165
https://doi.org/10.38124/ijisrt/IJISRT24APR227

lack of training, with plans to enhance the system in future

research. [18] The research evaluates 28 Twitter sentiment  Text Blob
analysis systems, highlighting challenges such as brevity, It is a library for natural language processing and
sentiment class imbalance, and stream-based generation and sentiment analysis. It simplifies text processing tasks and
poor performance (average accuracy 61%). Domain-specific offers sentiment analysis capabilities.
models and higher recall improve event detection. [19] The
study presents an advanced Twitter sentiment analysis  Word Cloud
model, combining word embeddings, n-grams, and It is used to create word clouds, which are visual
sentiment features in a deep convolutional neural network. representations of word frequencies. In this project, it's used
Outperforming baseline models, it demonstrates the efficacy to generate a word cloud from the tweet data.
of pre-trained word vectors. [20] This research proposes an  Pandas
ensemble classifier for Twitter Sentiment Analysis, It is a powerful data manipulation library in Python. It
combining base classifiers (Naive Bayes, Random Forest, is used for creating and manipulating data in tabular form.
SVMs, Logistic Regression) to enhance accuracy. The It's used to create a data frame to store and work with tweet
results demonstrate superior performance over standalone data in the project.
and majority voting classifiers, applicable for product
monitoring and consumer decision-making. [21]  NumPy
It is a library for numerical computations in Python.
While it's not explicitly used in this code, it's often used in
III. METHODOLOGY data analysis and manipulation tasks.

A. Libraries  Re
The program relies on the utilization of several It is Python's built-in library for regular expressions.
essential libraries to facilitate various tasks. These libraries It's used to perform text cleaning and preprocessing by
are as follows: removing specific patterns (e.g., mentions, hashtags,
hyperlinks) from the tweet text.
 Tweepy
It is a Python library for accessing the Twitter API. It  Matplotlib
provides functions and classes to interact with Twitter, such It is a library for data visualization. It provides
as fetching tweets, posting tweets, and managing user functions for creating various types of plots and charts. In
accounts. the project, it's used to visualize sentiment analysis results.

Fig 1: Flowchart Illustrating the Sequential Process for Conducting Sentiment Analysis on Twitter

B. Twitter API Interaction These credentials encompass the ‘consumerKey’ and

‘consumerSecret’. These application-level credentials are
 Twitter API Credentials crucial for the program to authenticate and gain access to the
To initiate interaction with the Twitter API, an API Twitter API programmatically. Once the authentication
object is instantiated, equipped with the necessary object is created and populated with these credentials, it sets
credentials obtained from the Twitter developers' forum. the user-specific credentials, ‘accessToken’ and

IJISRT24APR227 www.ijisrt.com
2068
Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research
Technology
ISSN No:-2456-2165
https://doi.org/10.38124/ijisrt/IJISRT24APR227

‘accessTokenSecret’, effectively forming an API object irrelevant or distracting information. The ‘apply’ method is
('API') through which the program communicates with then used to apply this function to the ‘Tweets’ column of
Twitter's API. This API object serves as the gateway for the data frame. making the text more suitable for analysis.
retrieving tweets from the specified Twitter account. The resultant cleaned tweets are stored in a ‘pandas’ data
frame denoted as 'df'. It has a single column labelled
 Creating the Authentication Object and Setting the ‘Tweets’, where each row contains the content of one tweet.
Access Token This data frame serves as an organized and structured
The ‘tweepy.OAuthHandler’ class is used to create an tabular representation of the tweet data, enhancing its
OAuth authentication handler called ‘authenticate’. This manipulability and analytical potential.
handler is configured with the ‘consumerKey’ and
‘consumerSecret’. It's a necessary step to access the Twitter
API securely. After creating the authentication object, the D. Sentiment Analysis
access token and access token secret are set using the
‘set_access_token’ method. This step associates the script  Subjectivity and Polarity Evaluation
with the specific Twitter account authorized through these The ‘TextBlob’ library is utilized to create two
credentials. essential functions: 'getSubjectivity' and 'getPolarity,' which
serve to evaluate the subjectivity and polarity of individual
 Creating the API Object tweets. Subjectivity is a metric that quantifies the extent to
The ‘tweepy.API’ class is employed to create an API which the text expresses opinions or objectives, whereas
object named ‘api’. This object allows interaction with the polarity is indicative of the sentiment contained within the
Twitter API using the authentication previously set up with text, classifying it as either positive, negative, or neutral.
‘authenticate’. The ‘wait_on_rate_limit=True’ parameter These two metrics are represented as numerical values to
means that the script will wait if it reaches Twitter's rate provide a clear understanding of their intensity.
limits when making API requests.
 Classification and Categorization
 Data Acquisition - Extracting Tweets Texts having scores of zero are labelled as neutral,
The user is requested to enter the Twitter username of those with scores greater than zero are classified as positive,
the account they want to analyse, along with the desired and those with scores below zero are characterized as
count of the most recent tweets to extract from the dataset. unfavourable or negative. Subsequently, the tweets that
In this case, the specified Twitter handle is that of the US exhibit the most pronounced positive and negative
President, Joe Biden, and the number of tweets to be inclinations are further identified and displayed. This
extracted is 50. The ‘api.user_timeline’ method is then categorization process is executed by leveraging the scores
employed to obtain the specified number of recent tweets in assigned to each tweet, enabling the separation of tweets
extended tweet mode and store them in the ‘posts’ variable. into their respective sentiment categories.

C. Data Preprocessing - Text Formatting and Data Frame To conduct sentiment analysis and categorize tweets as
Creation 'Positive,' 'Negative,' or 'Neutral,' the 'getAnalysis' function
The text from the extracted tweets undergoes is introduced. This categorization is determined by
formatting to remove extraneous elements such as considering the polarity score of each tweet. The outcomes
@mentions, symbols, and hyperlinks. This text-cleaning of the sentiment analysis are then incorporated into the data
process is executed by the 'cleanTxt' function, which frame, as illustrated in Fig. 2, as a new column labelled
prepares the text for sentiment analysis by eliminating any 'Analysis,' enabling the easy classification of each tweet.

IJISRT24APR227 www.ijisrt.com
2069
Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research
Technology
ISSN No:-2456-2165
https://doi.org/10.38124/ijisrt/IJISRT24APR227

Fig 2: Quantitative Sentiment Analysis of Twitter Content is Facilitated by Examining the Subjectivity and Polarity of the Tweets

Positive and negative tweets are separately displayed tweets and ‘ascending=False’ for negative tweets. A ‘for’
using the data frame. The 'df' data frame is arranged in order loop iterates through the sorted data frame. Inside the loop,
based on the 'Polarity' column. Tweets with positive it checks the ‘Analysis’ column to determine if a tweet is
sentiment have higher polarity values and tweets with positive or negative. If it's positive or negative, the tweet
negative sentiment have lower polarity values. This is done number and the tweet text are printed as seen in Fig. 3.
using ‘df.sort_values’ with ‘ascending=True’ for positive

Fig 3: Tweets Categorised as Positive and Negative amongst the Extracted Tweets

IJISRT24APR227 www.ijisrt.com
2070
Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research
Technology
ISSN No:-2456-2165
https://doi.org/10.38124/ijisrt/IJISRT24APR227

Using the data frame, the positive and negative tweets  Percentage
are printed separately. The data frame ‘df’ is sorted based on The data frame ‘df’ is filtered to create a new data
the ‘Polarity’ column. Tweets with positive sentiment have frame ‘ptweets’ and ‘ntweets’ that only contains positive
higher polarity values and tweets with negative sentiment and negative tweets respectively. The percentage is
have lower polarity values. This is done using calculated by dividing the count of positive tweets,
‘df.sort_values’ with ‘ascending=True’ for positive tweets represented by 'ptweets.shape[0],' and the count of negative
and ‘ascending=False’ for negative tweets. A ‘for’ loop tweets, denoted by 'ntweets.shape[0],' by the total number of
iterates through the sorted data frame. Inside the loop, it tweets in the dataset, 'df.shape[0],' individually. The result is
checks the ‘Analysis’ column to determine if a tweet is rounded to one decimal place using ‘round()’. This yields
positive or negative. If it's positive or negative, the tweet the percentage of positive and negative sentiments within
number and the tweet text are printed as seen in Fig. 3. the tweets, as illustrated in Fig. 4.

Fig 4: Percentage of Positive and Negative Tweets amongst the Extracted Tweets

IV. RESULTS AND DISCUSSION and its presentation is facilitated through the 'matplotlib'
library.
A word cloud serves as a visualization method
employed to represent textual information, with the The word cloud in Fig. 5, shows words related to
magnitude of each word mirroring its occurrence or politics and the US presidential election. It reveals the
importance. This technique visually portrays the words that political campaign and vision of the president, the social and
appear most often, with word size and colour reflecting their environmental issues and policies, the legislative and
frequency. Consequently, it offers a rapid snapshot of the administrative aspects, and the outreach and mobilization
prevalent terms within the dataset. The size of the word efforts of his political party.
cloud is established by utilizing the 'Word Cloud' library,

IJISRT24APR227 www.ijisrt.com
2071
Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research
Technology
ISSN No:-2456-2165
https://doi.org/10.38124/ijisrt/IJISRT24APR227

Fig 5: The Word Cloud is Generated to Visually Represent the Most Frequently used Words in the Tweets by the User,
Highlighting Significant Terms

A scatter plot is created to visualize the relationship As seen in Fig. 6, the resultant scatter plot indicates
between polarity and subjectivity for all the tweets. It uses that the tweets have a variety of sentiments. Some points are
‘plt.scatter’ to plot each tweet's polarity (x-axis) and close to the origin, meaning that they are neutral or factual.
subjectivity (y-axis). ‘plt.show()’ is called to display the Some points are far from the origin, meaning that they are
scatter plot. strongly positive or negative, or highly opinionated.

Fig 6: The Graph Depicts the Tweets' Subjectivity and Polarity Relative to Each Other

The 'matplotlib' library is employed to create a Fig. 7 illustrates the distribution of positive, neutral,
graphical representation of tweets based on their subjectivity and negative sentiments within the textual data through
and polarity. The ‘value_counts()’ function is used to obtain three distinct bars of a bar graph. The positive sentiment has
the counts of each sentiment category in the ‘Analysis’ the highest count, approximately 80, followed by the neutral
column, ‘plot(kind='bar')’ is used to create the bar chart, and and negative sentiments, both having fewer than 20 counts.
‘plot(kind='pie')’ is used to create the pie chart. This Also in Fig. 7, the generated pie chart instead has those three
visualization is enhanced by using the percentages of sections represented using distinct colours with the blue
positive and negative tweets. The outcomes are showcased (positive) section covering the majority of the chart
in a pie chart and a bar graph, employing unique color- followed by the orange (neutral) and yellow (negative)
coding to illustrate the distribution of positive, negative and sections. Both visualisations indicate that the tweets have a
neutral tweets. more favourable or optimistic expression than unfavourable
or pessimistic ones.

IJISRT24APR227 www.ijisrt.com
2072
Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research
Technology
ISSN No:-2456-2165
https://doi.org/10.38124/ijisrt/IJISRT24APR227

Fig 7: The Bar Graph and Pie Chart Visualization Illustrate the Distribution and Proportion of Counts for Neutral, Positive, and
Negative Tweets

V. FUTURE SCOPE
In this comprehensive research endeavour, a
The future of this sentiment analysis project holds sophisticated python application was developed for
significant potential for refinement and expansion. To sentiment analysis. It is designed to extract valuable insights
enhance user engagement, the implementation of an from the dynamic landscape of Twitter. It not only interacts
intuitive Graphical User Interface (GUI) would simplify the seamlessly with the Twitter API but also systematically
input procedure. This would enable users to effortlessly processes and presents data in an intelligible and visually
enter the Twitter username and designate the desired engaging manner.
quantity of latest tweets for analysis. Such user-friendly
accessibility would broaden the tool's utility to users with The program's capabilities include gathering recent and
varying levels of technical expertise. While the current pertinent tweets from a specific user and presenting them in
system offers compelling visualizations like word clouds, an organized and reader-friendly format. Advanced text
bar graphs, and pie charts, potential avenues include mining techniques are harnessed to evaluate the subjectivity
heatmaps, trend graphs, and sentiment over time analysis. and polarity of these tweets, allowing for a nuanced
These additions would provide more in-depth insights into understanding of the emotional content within the text. One
the temporal dynamics of sentiment within a user's Twitter of the key features of the program is the creation of a word
activity. cloud, a visually intuitive representation of the most
frequently used words by the user. This tool provides an
Incorporating machine learning and deep learning immediate grasp of the user's recurring themes and interests.
techniques has the potential to improve the accuracy of Additionally, the program identifies and highlights the most
sentiment classification. Additionally, exploring context- positive and negative tweets, shedding light on the user's
aware sentiment analysis could enable a more nuanced emotional engagement and further enriching the analysis.
understanding of language, including sarcasm. Real-time
analysis of Twitter data would enable users to monitor The program also calculates the percentage of positive,
evolving sentiment trends and respond proactively to negative, and neutral tweets within the user's content. This
emerging discussions. Furthermore, multilingual support, data is then visualized through informative bar graphs and
customizable sentiment categories, predictive analytics, and pie charts, offering a clear and compelling depiction of the
the integration of sentiment analysis from various social distribution of sentiment in the analysed tweets. By applying
media platforms are all areas that offer substantial potential text mining techniques and leveraging the capabilities of
for future development, making this project a dynamic and Python, the program uncovers and categorizes the emotions
evolving tool with far-reaching applications. expressed in tweets as neutral, positive, or negative.

The data generated through the sentiment analysis

program offers valuable insights into the collective
VI. CONCLUSION perspectives and sentiments of Twitter users. Such insights

IJISRT24APR227 www.ijisrt.com
2073
Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research
Technology
ISSN No:-2456-2165
https://doi.org/10.38124/ijisrt/IJISRT24APR227

are a powerful resource for enhancing decision-making analysis on worldwide COVID-19 outbreaks."
processes, from businesses adapting to customer sentiment Kurdistan Journal of Applied Research (2020): 54-
to policymakers understanding public concerns and 65.
sentiments. In essence, this would facilitate the efficient [11]. Asghar, Muhammad Zubair, Aurangzeb Khan,
analysis of social media data and contribute to the broader Furqan Khan, and Fazal Masud Kundi. "RIFT: a rule
goals of societal betterment by promoting informed and induction framework for Twitter sentiment analysis."
empathetic responses to the needs and concerns of the Arabian Journal for Science and Engineering 43, no.
public. 2 (2018): 857-877.
[12]. Wang, Lei, Jianwei Niu, and Shui Yu. "SentiDiff:
REFERENCES combining textual information and sentiment
diffusion patterns for Twitter sentiment analysis."
[1]. Vadai, Gergely, Zoltan Gingl, and Robert Mingesz. IEEE Transactions on Knowledge and Data
"Generalized attack protection in the Kirchhoff-Law- Engineering 32, no. 10 (2019): 2026-2039.
Johnson-Noise secure key exchanger." IEEE Access [13]. Dubey, Akash Dutt. "Twitter sentiment analysis
4 (2016): 1141-1147. during COVID-19 outbreak." Available at SSRN
[2]. Rehioui, Hajar, and Abdellah Idrissi. "New clustering 3572023 (2020).
algorithms for twitter sentiment analysis." IEEE [14]. Ghiassi, Manoochehr, and Sean Lee. "A domain
Systems Journal 14, no. 1 (2019): 530-537. transferable lexicon set for Twitter sentiment analysis
[3]. Symeonidis, Symeon, Dimitrios Effrosynidis, and using a supervised machine learning approach."
Avi Arampatzis. "A comparative evaluation of pre- Expert Systems with Applications 106 (2018): 197-
processing techniques and their interactions for 216.
twitter sentiment analysis." Expert Systems with [15]. Saad, Shihab Elbagir, and Jing Yang. "Twitter
Applications 110 (2018): 298-310. sentiment analysis based on ordinal regression."
[4]. Nuser, Maryam, et al. “Sentiment Analysis of IEEE Access 7 (2019): 163677-163685.
COVID-19 Vaccine Using Deep Learning.” Journal [16]. Wazery, Yaser Maher, Hager Saleh Mohammed, and
of Theoretical and Applied Information Technology, Essam Halim Houssein. "Twitter sentiment analysis
vol. 100, no. 12, 30 June 2022, using deep neural network." In 2018 14th
www.jatit.org/volumes/Vol100No12/19Vol100No12. international computer engineering conference
pdf. (ICENCO), pp. 177-182. IEEE, 2018.
[5]. Alharbi, Ahmed Sulaiman M., and Elise de Doncker. [17]. Asghar, Muhammad Zubair, Fazal Masud Kundi,
"Twitter sentiment analysis with a deep neural Shakeel Ahmad, Aurangzeb Khan, and Furqan Khan.
network: An enhanced approach using user "T‐SAF: Twitter sentiment analysis framework using
behavioural information." Cognitive Systems a hybrid classification scheme." Expert Systems 35,
Research 54 (2019): 50-61. no. 1 (2018): e12233.
[6]. Guo, Xinyi, and Jinfeng Li. "A novel twitter [18]. Elbagir, Shihab, and Jing Yang. "Twitter sentiment
sentiment analysis model with baseline correlation analysis using natural language toolkit and VADER
for financial market prediction with improved sentiment." In Proceedings of the international
efficiency." In 2019 Sixth International Conference multiconference of engineers and computer
on Social Networks Analysis, Management and scientists, vol. 122, no. 16. sn, 2019.
Security (SNAMS), pp. 472-477. IEEE, 2019. [19]. Zimbra, David, Ahmed Abbasi, Daniel Zeng, and
[7]. Naseem, Usman, Imran Razzak, Katarzyna Musial, Hsinchun Chen. "The state-of-the-art in Twitter
and Muhammad Imran. "Transformer based deep sentiment analysis: A review and benchmark
intelligent contextual embedding for twitter evaluation." ACM Transactions on Management
sentiment analysis." Future Generation Computer Information Systems (TMIS) 9, no. 2 (2018): 1-29.
Systems 113 (2020): 58-69. [20]. Jianqiang, Zhao, Gui Xiaolin, and Zhang Xuejun.
[8]. Zhang, Yazhou, Dawei Song, Peng Zhang, Xiang Li, "Deep convolution neural networks for twitter
and Panpan Wang. "A quantum-inspired sentiment sentiment analysis." IEEE access 6 (2018): 23253-
representation model for twitter sentiment analysis." 23260.
Applied Intelligence 49 (2019): 3093-3108. [21]. Saleena, Nabizath. "An ensemble classification
[9]. Joshi, Shaunak, and Deepali Deshpande. “Twitter system for twitter sentiment analysis." Procedia
Sentiment Analysis System.” International Journal of computer science 132 (2018): 937-946.
Computer Applications, vol. 180, no. 47, 15 June
2018, pp. 35–39,
https://doi.org/10.5120/ijca2018917319.
[10]. Manguri, Kamaran H., Rebaz N. Ramadhan, and
Pshko R. Mohammed Amin. "Twitter sentiment

IJISRT24APR227 www.ijisrt.com
2074
Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research
Technology
ISSN No:-2456-2165
https://doi.org/10.38124/ijisrt/IJISRT24APR227

IJISRT24APR227 www.ijisrt.com
2075