Volume 9, Issue 1, January 2024 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Sentiment Analysis for Enhancing Business Process

using Naive Bayes
Emmanuel G. Galupo Jr.; Jeffrey F. Calim; Emmie Faye Marione L. Matabile; Johani D. Basaula;
Luisa M. Mariano; Arnel Balasta; Aerob C. Robles; Antoniette C. Mariano
Our Lady of Fatima University College of Computer Studies Quezon City

Abstract:- In order to grow, businesses nowadays process reviews through the use of machine learning and
must ob- tain customer feedback, such as reviews sentiment analysis[1]. A well-designed system greatly
or comments. They are thereby collecting additional reduces the possibility of bias when using client input.
information. The process of manually collecting and Sentiment analysis is a rapidly developing field in natural
analyzing data is becoming more and more onerous language processing that may be applied in business to
for the owners of these changes. The purpose of this quickly un- derstand people’s opinions, attitudes, and
research is to develop an algorithm-based system that feelings about a product by looking through a large amount
can au- tomatically extract data and support business of data. The majority of people, who are usually
activities. The tech- nology will reduce the effort of individuals, voice their opinions about various objects or
human workers in data analysis because it will topics. These topics usually include the products or
automatically examine the entered data. It features a services of a certain company that have received reviews.
sentiment analysis graph. It also offers a word cloud Sentiment analysis also aims to identify the emotion
that makes things easy to comprehend for the business conveyed in the text, extract relevant information, and
administrator by displaying the most relevant analyze it. Business analysis is a talent that experts in
keyword in different sizes according to how business must have. On the other hand, predictive
frequently the system identified the word from analysis will examine the model to ascertain the best way to
reviews or collected data. The system will forecast do a specific task. Sentiment analysis can be assessed at
which department or sector of the enterprise needs three different levels: aspect, phrase, and document levels.
improvement. The researchers will build the system Finding out if a document conveys a positive or
using the Iterative System Design Life Cycle since negative entity is the aim of document-level analysis.
it is most equipped for handling erratic behavioral Sentence-level analysis is performed to determine whether
shifts and even data science. Using brainstorming the sentiment expressed in each sentence is an opinion.
approaches, the project concept and approach for This stage seeks to ascertain if the statement offers
this study were explored or written. The opinions or accurate information. Aspect- level analysis
instruments for requirements formulation, such as does a great job of classifying sentiment toward entity
customer interviews and system functionality, aspects. At this stage, identify the entity and its parts. The
usability, and security assessments, must be chosen idea of artificial intelligence’s machine learning is to
by the researchers. develop a computer program that can learn only from
data and without human assistance[9]. There are three types
Keywords:- Sentiment Analysis, Machine Learning, of it, and each one is a difficult topic in the field of
Feed- backs, Sentiment, Multinomial Naı̈ve Bayes. information technology: supervised learning involves
labeling the training data and defining both the input and the
I. INTRODUCTION output in order to find patterns and correlations.
Unsupervised learning use unlabeled data—which only
In order to understand how to improve the company characterizes the input—to discover patterns and significant
ser- vices, one of the most important phases in the processes relationships. Reinforcement learning employs
of most businesses and other e-commerce firms is to predetermined rules and entails doing an action that,
acquire information from the target customers, such as depending on the outcome, will either result in a positive or
feedback, comments, and opinions. As time goes on, the negative response from the machine [10]. It solves real-
amount of data that must be gathered grows faster, world problems by employing a variety of computer
necessitating a shift in the methods used to collect and approaches to create models using datasets—collections of
organize it. Furthermore, as this generation is already data that are examined in order to produce the model.Once
engaged with the ”tech world,” more recent methods like built, the model may now perform actions, make decisions,
sentiment analysis are being employed [2]. Companies and generate forecasts based on the provided dataset.
may have mountains of customer feedback gathered, but in Despite the rapid advancement of technology, several
today’s world, organizations are suffering from data fatigue businesses continue to manually organize their data using
(which doesn’t imply better or deeper insights). But for us Excel and paper. Additionally, some people will never be
ordinary people, manually analyzing data without prejudice able to recover from an unplanned data loss that could
or inaccuracy is never easy. Due to this problem, the occur to them at any time using that kind of technique,
researchers developed a system that enhances business which means the analysis will be disregarded and lose all

IJISRT24JAN1363 www.ijisrt.com 2374

Volume 9, Issue 1, January 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
relevance. Running a business requires a lot of labor, customers. The data that the technology will collect will
particularly with regard to data. Nowadays, sentiment help the company pinpoint its issues. All of the data that
analysis may be used by businesses to rapidly determine has been examined will be shown in the dashboard
the polarity of any sentiment, be it an automated review, following a user’s submission of feedback that is
public comment, opinion, or customer feedback. automatically processed by the system and uploaded to
Additionally, they can employ sentiment analysis to deliver the database. In order to assist the company in identifying
insights or analytics that help business owners make common input—whether positive, negative, or neutral—the
decisions that will further enhance their company. This is system will also indicate words that are used a lot.
made possible via machine learning and sentiment Additionally, the system will display all of the data that
analysis. Whether a feeling is favorable, negative, or has been researched, emphasizing the features of the good
neutral, human analysts who assess sentiments based on or service that the company needs to improve. As long
research tend to agree on its polarity in 80–85Evaluating as the data is collected by the company and approved
sen- timent analysis models using a Naive Bayes for use in this project, the system is limited to predicting
algorithm—more especially, Multinomial Naive Bayes—is the polarity of data—positive, negative, or neutral—
the aim of the study. This study uses a variety of originating from the dataset about customers’ sentiments.
sentiment analysis approaches in an effort to improve the This includes opinions, comments, feedback, reviews, and
business process based on end- user assessments or surveys. A word cloud that displays the most relevant
feedback. The technology analyzes the data using a term and its text size based on how many times it
document-level algorithm after automatically extracting it. appears in the dataset will be created by extracting and
Using models, the system predicts the sentiment at the analyzing all the data, including the recently anticipated
document level. As a result, the entire review of a data that has been added to the primary data store. This can
person is used as the data to create a clean version of assist the business create a plan and provide insight into the
it, which is then analyzed by the model and free of words most common comments made about it. Based on the data
like is and similar keywords. It’s challenging to run a the system analyzes, a firm may be able to decide how to
corporation or business. Many adjustments, tasks, and improve its offerings by displaying the aspects of a specific
endeavors must be completed in order to make things right. product or service that could be improved. Furthermore, the
Taking into account that the business must meet the needs of system is limited to analyzing spoken English. The only
its clients. As a result, businesses are accepting more and place to create an account is on the login page. The first
more consumer input, which could potentially be difficult account to be created will come with an admin role by
to manage, particularly if there are a lot of comments. default; however, later accounts that are created will not
Therefore, having a system that could help a business will have one and will need to be granted one by the admin in
make receiving customer feedback easier. a system that order to be able to access the system. Additionally, the
keeps track of, evaluates, and filters the issue the business is system restricts account access to the following three roles:
facing. After analyzing the comments, the system will Admin, Staff, and Guest. While the staff job can only view
automatically highlight the terms that customers used most the dashboard, predict, train, and profile pages, the admin
frequently. The dashboard would display all of the data position can access all pages, including the admin page. The
that was analyzed, which might help the company find dashboard page is the sole page accessible to guests. The file
solutions. The objective of this project is to create a input on the anticipated page can only have one column—the
system that may enhance business process assessments of review for each row—and must also have the file type
the company and lessen the workload related to data extension.csv. The dataset that needs to be imported must
analysis. The system automatically analyzes the data that is have a file extension of.csv and can only have two columns:
submitted, and it shows the information and sentiment polarity in the second column and a review or
analysis graph so that the business may assess the sentiment in the first for each row. The following
aspects of the reviews for possible modifications in the things have not been included in this project: gathering
future. The system also has a word cloud that shows the information, especially customer reviews of the company;
most essential terms in different sizes based on how creating suggestions or ideas for new ideas to progress
frequently the system identified them in the assessments, the business (the system can only comprehend English at
making it easy for the company administrator to rapidly the moment; this could change based on client data);
identify the most important words. Technology will help displaying, modifying, and retrieving the password; and
businesses by helping them identify where modifications notifying users when a new account is created. the process
need to be made and by anticipating the polarity of a of making a new account for any role using an admin
customer’s attitude or review. The company will benefit account.
from knowing end users’ preferences for certain goods or
services. The primary objective of this initiative is customer II. RELATED WORK
feedback for the company. The business will use this
project’s machine learning-based sentiment analysis to The growth of e-commerce has increased the
automatically produce and analyze customer feedback competition between the new players and the traditional
once it has received data from the firm. All of the data that offline entrants. As a result, social media platforms are
is analyzed and helps the company discover the problem introducing the business world to a more practical means of
will be generated by the system. This project will be expressing one’s thoughts or emotions regarding a certain
centered around the input that the firm will get from its service within the sector. Because of this, end consumers

IJISRT24JAN1363 www.ijisrt.com 2375

Volume 9, Issue 1, January 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
frequently inquire about the thoughts and opinions of others utilized one in business analytics, and it highlights the
regarding particular products or services. As a result, importance of incorporating machine learning into business
sentiment analysis is becoming more and more important analytics. This highlights the application areas of popular
for organizations to help address the lack of certain services machine learning algorithms for business analytics and
by creating improvement initiatives. Electronic word-of- introduces their significance [8]. They employed a
mouth, or E-WOM, is utilized in marketing to inform random forest classifier method due to its proficiency in
decisions [2]. Electronic word- of-mouth has been predictive analysis. It also provides an attribute that
employed as a method for marketing research, and evaluates how crucial it is to train a model, like
businesses can use social media to get customer feedback. extracting insights. The benefit of this approach is that it
Sentiment analysis combined with eWOM can help has suitable sample size and dimensionality scalability,
analyze a particular issue and provide solutions for the yields results that are useful, and requires minimal pre-
company. This study’s format ought to encompass the processing of the data. Based on the baseline random
application portion as well as the theoretical performance value of 0.5, which is equal to 50% of the
underpinnings of associated ideas. Electronic word-of- sentiment analysis model that was produced, the algorithm’s
mouth is included in the theoretical basis, and specific performance is not very excellent. Studies that can forecast
studies and data are used in the sentiment analysis. a company sector and provide an explanation based on
Three-stage designs were utilized in the application part financial statements have employed machine learning.
for sentiment analysis. Three distinct scenarios, comprising Financial statements are now contained in a sizable
an algorithm and a practical application, are included in collection of data that is made accessible through open data
each step. These nine scenarios imply that electronic word- sets. Prior to now, the primary goal of dealing with such
of-mouth comprehension is meant to be used in sentiment data was to forecast fraud and bankruptcy. However, we can
evaluation [3]. Businesses can benefit from machine now produce a predictive version that can help achieve
learning (ML) in the management and utilization of the same datasets with a missing information sector. The
user-generated data. In order to create an ontology based proposed methodology treats business sector prediction as a
on the following phases, businesses need machine learning classification task by means of a supervised learning
(ML): (1) describing data problems, (2) building approach based on random data. Upon assessing the
solutions centered on fundamental causes, (3) understanding attribute deemed crucial for the ultimate classification
use cases, and (4) creating ontology organization principles model, we noticed that certain feature scales are utilised for
[5][6]. ”Its algorithms can still learn to perform tasks forecasting significant business sectors. Furthermore, an
without being given clear instructions.” [13] As an attribute’s existence or absence depended on something
illustration, it can be used to predict the likelihood that other than its inherent value. The effect of the insight can be
complex production equipment will break down. One demonstrated in accounting, where it is investigated how
method for evaluating a client’s assessment from derived company features and financial statements relate to one
tweets is natural language processing. Views are seen as another [9]. Sentiment analysis extracts information from
outcomes that need to be quantified. This study looks at huge data using text and natural language. These days,
tweets from the beginning of the lockdown in the Luzon people use social media to share their opinions in a variety
region to the third week. It automatically classifies of ways, such as by leaving comments on particular
mentions of the coronavirus and Covid-19 on social media articles, reviewing products or services, or occasionally
according to sentiment analysis. Sentiment analysis is submitting their own content on these platforms. Sentiment
used to help specialists make decisions based on their analysis of data has demonstrated that it can have a
emotions. Sentiment research thus reveals that the excessive significant influence on the decisions made to enhance the
party quarantine has multiple effects on the majority of government and specific enterprises. Owners of
Filipinos. According to popular belief, one’s basic businesses occasionally use user or customer reviews that
needs—namely, access to food and government funding— have been placed to assess issues and improve their
are in danger. While some are considering the positive services so that they are suitable for their customers
aspects of COVID-19, it was mistakenly believed that the [14]. The majority of web applications are written in
virus was to blame rather than the actions of Luzon java,.net, or other web languages other than Python, which
Twitter users who implemented lockdowns, group makes it extremely difficult to construct machine learning
quarantines, and social distance. It was anticipated that there using the Python ML technique. When a web
would be a surge in negative sentiments among Twitter application written in Java is asked to perform a
users, as negative sentiments are known to rise with time prediction or train a model, for instance, it must
[7]. Numerous analytical techniques have long been communicate with the Python machine learning algorithm
employed to address business-related inquiries through the Java thread. In order to get around this obstacle,
concerning the primary methodology of data warehouse Django is a web-based technology that is entirely written in
research. Most significantly, data mining and business Python and eliminates the need for cross-language
intelligence help decision makers obtain the information communication because the ML algorithm it uses is also
they need. Numerous systems are attempting to adjust to written in Python, which can increase its efficiency [15].
the new regulations that have been modified in the One of the difficulties with sentiment analysis is the
past few years. There is no denying big data’s allure. language barrier because sentiment can be expressed in a
In particular, the problem of assessing large amounts of variety of languages, including Urdu, English, and many
data derived from social media is the most frequently more. The majority of sentiment analysis research are

IJISRT24JAN1363 www.ijisrt.com 2376

Volume 9, Issue 1, January 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
written in English, and they use WEKA (Waikato frequency or other criteria, and may include graphic
Environment for Knowledge Analysis) to classify text in elements such as font, color, and size [27]. Its value
Roman Urdu in order to construct a sentiment analysis for lies in screening, analyzing, and comprehending seemingly
reviews of cars in that language [16]. A study jumbled textual data; it also exposes key concepts by the
comprising 2000 reviews—1000 positive and 1000 new pattern and provides the text’s focus. It highlights
negative—from several car websites was utilized to test the most significant terms in a clear and quantifiable
various algorithms. The study’s findings indicate that the manner. By enabling readers to rapidly assess and
deep neural network classifier has the highest accuracy examine the text, even across a range of texts, a word
result at 82%, followed by the decision tree at 75.75%, cloud can assist readers in understanding the theme and
bagging at 84.5%, random forests at 78.75%, k-NN at connotation from a number of angles. The suggested
72%, AdaBoost at 83.75%, SVM at 76.5%, and multi- model includes comparisons, and the outcome implies that
nomial Naive Bayes at 89.75%. Machine learning-based stacking ensembles improves prediction accuracy. Support
sentiment analysis is concerned with classifying reviews as Vector Machine was determined to be more accurate overall
either good or negative, and it is inferred through the than the other two, while K-Nearest Neighbor is
extraction of thoughts, emotions, text subjectivity, and substantially less accurate than Naive Bayes and Support
other elements [26]. In addition, the machine picks up Vector Machine. About 20% more precision has been
the ability to appropriately assess emotion on its own obtained [22]. Two often used techniques for estimating
without human oversight [11]. Since sentiment analysis may and optimizing sentiment are Support Vector Machines
be automated to the point that decisions can be made based (SVM) and Naive Bayes classifiers. Hierarchical machine
on a large quantity of data rather than simply gut learning techniques do ok in classification problems,
instinct, which isn’t always correct, it will help determine however SVM and Multinomial Naive Bayes have proven to
what the most pressing issues are at this time. Sentiment be effective in terms of precision and optimization.
analysis is a study method that helps researchers fully Sentiment analysis using neural network designs has
understand how consumers make decisions by analyzing surfaced in a few sentences [23]. The sentiment prediction
and identifying the feelings and opinions of users or methods that use deep convolutional neural networks and
customers [17]. Businesses are beginning to employ recursive neural networks are a little more involved in
sentiment analysis as a result of recent changes to their terms of capturing the semantics of phrases. Convolutional
business models and methods, which include social neural networks were able to derive word- or sentence-level
developments and processes that can help them adapt to qualities like stems and morphological tags, however many
the current world. makes use of Amazon reviews to neural network architectures found it difficult to extract
ascertain the neutrality, positivity, and negativity of the character-level features and embeddings of abstract
comments. Additionally, this study examined the two words. Few researchers have employed J48, BFTree, and
machine learning algorithms—the Naive Bayes method OneR for sentiment estimation. On Twitter, these three
and the Support Vector Machine (SVM)—that will underpin classifiers are utilized for text categorization and text
the sentiment analysis model as a whole by classifying emotion detection. Sentiment analysis is becoming more
client input into three groups [19]. The imputed dataset will and more popular, particularly in the corporate world. With
compute the positive, neutral, and negative categorization the right methods, techniques, and tools for sentiment
scores using Sentiwordnet. The evaluation is contingent analysis concerning businesses, it will provide a solution for
upon the presentation and can be ascertained by applying decision-making processes to gain some profits [24]. It also
the F-1 Measure, Accuracy, Precision, and Recall to each has some benefits in any aspect of business practice, such as
classification. Based on the exploratory results, Naive review of various product and/or service feedbacks, and
Bayes classification outperforms Support Vector Machine so on. Sentiment analysis can be applied to a wide range
(SVM) in terms of accuracy. True positive examples (TP), of business endeavours or establishments, including blogs,
False positive examples (FP), True negative examples reviews of books and movies, comments about restaurants,
(TN), and False negative examples (FN) were used to and assessments of products and services [25]. In this
finish the figures. A sentiment identification classifier case, the company is able to observe customer opinions
can be made using Deep Learning, sometimes referred to about them, which they can capitalize on to enhance their
as Deep Neural Network, a machine learning component. A offerings. The three main categories of sentiment analysis
few examples of Deep Learning are CNNs (Convolutional approaches currently in use are machine learning, lexicon-
Neural Networks), RNNs (Recurrent Neural Networks), based, and hybrid approaches. A hybrid algorithm is the
and DBNs (Deep Belief Networks). Researchers have most effective methodology for this type of study since it
been motivated by deep learning approaches because of combines deep learning and machine learning to perform
their potential performance in comparison to traditional better at sentiment categorization [26]. Additionally, the
methods like Naive Bayes (NB) and Support Vector hybrid approach can comprehend sentiment classification
Machine (SVM) [20]. According to reports, the World in both bilingual and single-language texts. For the
Cloud is reportedly utilized in a number of contexts to purpose of independent research, models such as word
generate diagrams by cleaning text in the phrases that vectors, bag-of-words, Naive Bayes Support Vector
appear most frequently. This is typically accomplished on a Machine (NB-SVM), and recurrent neural networks (RNNs)
regular basis as a plain text outline [21]. Word clouds, often with long short-term memory (LSTM) are examined
called text clouds or tag clouds, are visual in the sentiment classification job. As a consequence,
representations of keywords arranged according to word the analysis demonstrates excellent accuracy with an 89%

IJISRT24JAN1363 www.ijisrt.com 2377

Volume 9, Issue 1, January 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
hybrid method result, demonstrating that it can be paradigm, and statistical tools that analyze the collected
independently compared to the other technique. Three (3) data.
different types of issues that machine learning (ML) can
address are among the many factors to take into account A. System Design
when explaining ML [4]: The Iterative System Design Life Cycle was adopted
 Classification: by recognizing objects and by the researchers to build the system since it is the most
comprehending text or speech.Classification suitable method for handling unanticipated changes in
incorporates recommenda- tions and correlations behavior and even in data science. In the planning stage, the
through cluster segmentation. researchers began to strategize about determining the
 Prediction/Estimation: the ability to anticipate and project’s viability and, concurrently, the potential results
project future events. once development got underway. In order to finish the
project, this process frequently includes gathering materials
ML ”can produce content in this scenario, from from both humans and machines. During the analysis phase,
interpolating missing data to generating the next frame in the researchers looked at the system’s database, flow, and
a video sequence,” for example. structure. This procedure allowed the system to operate
rationally. To provide the project a visual representation,
every concept that was reviewed during the design phase
III. METHODOLOGY has been advanced to the next level. Additionally, this is
where the researchers assemble the tools needed to develop
The researchers present the structure of the a system. Making sure the system functions well and
methodology that has been used in the study. This study accomplishes its objective is the notion behind the earlier
contains method- ology, algorithm, system design or steps used in the implementation phase.
paradigm, stages or phases under system design or

Fig. 1: Iterative System Development Life Cycle

 Planning: The planning stage of this study’s creation client to learn about the existing status of their firm.
involved the researchers researching current issues with They next looked at the business flow and processes.
sentiment analysis as well as its problems and To get a better understanding of the situation’s facts,
prospective remedies. As a result, the researchers the analysts also kept an eye on the business operations.
developed the topic “Sentiment Analysis using The analysts then examine the problem in order to
Machine Learning: A System to enhance business develop a solution that addresses the issue and a great
process reviews of the Company” and then gathered vision for the improvements the firm would like to see.
data from any companies that had feedback or This is done after monitoring the company’s current
reviews that were pertinent to this study. The main situation. The researchers will now go on to the length
problem with this topic is that humans must manually analysis, which necessitates a thorough investigation of
assess the polarity of sentiment; therefore, the how long each step in the creation of this project takes.
researchers intend to simplify it for people by creating a To produce a clear representation of the data flow in the
system. With the assistance of their colleagues and system, analysts must assess the data procedure.
consultants, the researchers created a system that  Design: The system’s architectural components
serves as a solution for this analysis. In order to have were identified by the researchers during the design
a success rate in sentiment prediction, the project also process in order to build the system’s architecture
employed the most effective and appropriate algorithm. design, which describes the hardware, software, and
 Analysis: The analysis phase is where the gathering network environment. The next item that the
of requirements is done to decide which resources to researchers created after creating the architecture
use in the requirements definition process. In order design was a flowchart of the system that shows the
to build this project, the analysts first interviewed the flow of algorithms and the sequential order of

IJISRT24JAN1363 www.ijisrt.com 2378

Volume 9, Issue 1, January 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
actions. In order to assure the interaction between
people and computers, the user interface design is a
component of the design process in the development
of this study. It aids in making this system easier
for people to use. In order to highlight development
requirements and show how the ultimate structure
can function, physical process models are created
during the design phase. The functionality of the
system in terms of data handling was described by
the physical data flow diagram. To create a theoretical
framework, the researchers gathered the
interconnected study themes. A conceptual framework
can help this study understand or visualize how
different variables interact with one another. This study
was based on sentiment analysis, a technology used
in modern enterprises for business analytics [24]. Fig. 2: Conceptual Framework of Sentiment Analysis
Nowadays, people will actually use an online Review
platform to express their opinions about the goods
they have used, the food they have consumed, the Jupyter Notebook as the Integrated Development
movies they have seen, the books they have read, and Environment (IDE) and Python 3 as the main
many other things that can be used as reviews and can programming language. The optimal algorithm for this
help to offer solutions when it comes time to make a analysis was selected by the researchers utilizing internet
decision [25]. Sentiment analysis has a wide range of data sets. The researchers created an interactive system
applications, and companies nowadays utilize it to get design on the Graphical User Interface (GUI) using the
client feedback. When processing all of the Hypertext Markup Language (HTML), Cascading Style
consumer feedback, machine learning is a business Sheets (CSS), Django framework, and Visual Studio Code.
algorithm that can be quite helpful. Machine learning The researchers used WampServer and MySQL for the back
can be used to analyze the typical positives and end. According to what was anticipated and accomplished
negatives of a client to create predictions and come to a during the design phase, the system’s creation began during
conclusion. As a result, employing sentiment analysis this time.
in company feedback with the use of machine
learning as an algorithm can now help generate and B. Algorithm
evaluate all the same problems at once, identifying In order to improve the system, the machine learning
their defects and certain things that require algorithm started to evolve. after some investigation and
improvement. The researchers can create a system that testing using a dataset of the same kind as the one used
genuinely helps a business owner better their service or in this work that is available online. The stacking classifier
product by gathering feedback that aids them in from the sci-kit-learn framework was originally going to
decision-making[9] by utilizing sentiment analysis be used by the researchers since it offers the best
and machine learning as an algorithm. accuracy among the algorithms they examined.
Additionally, a study reveals that a stacking technique that
Figure 2 illustrates how customer review data is employs K-Nearest Neighbor, Naive Bayes, and Support
entered into the system and, logically speaking, how the Vector Machine as base estimators and Support Vector
data is cleaned up inside the system using spaCy Machine as the final estimator could boost accuracy by
lemmatization, which reduces words to their simplest twenty percent (20%). Utilizing stacking algorithms, this
forms and eliminates words known as stop words that research will move forward. The researchers discovered that
have no bearing on the meaning of the sentence. The the system’s slow processing was a drawback for the
data is preprocessed after cleaning it so that the businesses that employed it. As a result, they made the
algorithm can comprehend it. The data was decision to switch to a more effective algorithm called
preprocessed and then subjected to sentiment analysis Naive Bayes, which has an accuracy that is
with the Naive Bayes technique to determine the comparable to that of the stacking algorithm and faster
polarity of the data. creation of a model to be used in predicting the
 Implementation: Programming tasks were carried out at review. Additionally, a study comparing various
this phase in order to create the concept for the algorithms, including Decision Tree, Bagging, Random
evaluated and planned system structure. To assure Forests, k-NN, AdaBoost, SVM, and Multinomial Naive
machine learning performance, the researchers carried Bayes, to be utilized in sentiment analysis revealed that
out experiments using Multinomial Naive Bayes produced a superior accuracy than
the others. To determine which class a new inputted piece
of data belongs to, Naive Bayes calculates the
likelihood that each feature in each piece of data in the
dataset will occur based on its class.

IJISRT24JAN1363 www.ijisrt.com 2379

Volume 9, Issue 1, January 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
C. Population and Sampling weighted mean and worded clearly and easily understood.
Because not every member of the population has an The standard formula or statistical tool is used to carefully
equal opportunity to be a sample or a respondent, the assess and handle the data collected by the researchers. The
researchers utilized convenience sampling, also known as average of all numbers is calculated by adding all the
opportunity sampling, in this study. A form of bias in numbers together and dividing the result by the total number
the selection of the respondents is non-probability of numbers. It is a statistical measure of the central
sampling. In order to include the greatest number of tendency of a probability distribution.
participants in this study, the researchers use a technique
called convenience or incidental sampling. Three (3) IT Table 1: Likert Scale
professionals who met the following criteria made up the Ratings Interpretation
respondents for the researchers: male or female; 21–50 3.51 – 4.00 Strongly Agree
years old; IT degree holder or IT professional; Filipino 2.51 – 3.50 Agree
citizen or foreigner; and eager to participate. Additionally,
1.51 – 2.50 Disagree
three (3) end users who work in management are chosen
based on a variety of characteristics, including gender, 1.0 – 1.50 Strongly Disagree
age (must be between 21 and 50), nationality (must be
either Filipino or foreign), and willingness to participate. The researchers also used the calculation of
There are six (6) respondents total, all of whom must percentages with the respondents’ surveys. The formula to
reside in or near the Philippines. get the 0% to
100% is:
D. Data Gathering Procedure
The researchers employed data-gathering Step value = 100/(categories -1) Step value = 100 / (4- 1)
techniques to learn more about the problem they
are studying. Through observation and survey questions, Step value = 33.33%
the study’s data are gathered. A variety of inquiries about
the project’s effectiveness were included in the survey. The calculation of percentage needs to calculate first
Additionally, instructions on how to evaluate the the maximum possible product, and this is the formula:
developed system and questionnaire were sent to the
responders. The researchers gave the respondents enough Maximum possible points = Total number of responses x
time to use the system and complete the survey Maximum percent
questionnaires. In this study, the researchers contacted the
chosen respondents by email, Facebook Messenger, and Maximum possible points = 9 x 100% Maximum possible
other social media platforms to see if they satisfied all points = 900%
the requirements for the population and sampling. If they
did, they were immediately eligible. The researcher gives The formula to get the percentage of the arithmetic
the qualified respondent an information sheet and consent mean is:
form so they can learn more about the study and
decide whether or not to participate on their own Percentage = Sum of the Products / Maximum Possible
volition. The individual can start the trials once they
have signed the consent form. The participant in the study Points x 100 / 1 Percentage = 866.67 / 900 x 100 / 1
examined the system for a maximum of thirty (30) minutes
before answering a questionnaire. The trial cycled three (3) Percentage = 96.30%
times. The procedure was carried out online utilizing
Remote Desktop and Google Forms for the qualified F. Research Ethics
participants who signed the consent form. Following the The researchers are looking for IT experts who can
system testing, the participant completes the questionnaire fulfill a number of criteria, including being male or
using a Google form. The participant first tests the female, between the ages of 21 and 50, possessing an IT
system via Remote Desktop, which enables the participant degree or working as an IT professional, being a Filipino
to test the researcher’s system. citizen or a foreigner, being able to read and write, and
being ready to participate. Additionally, management
E. Statistical Tool workers that are end-users have various selection criteria:
To collect the study’s data, the researchers prepared Male or female; between the ages of 21 and 50; employed
survey forms or questionnaires. The 4-point Likert Scale by the client’s company; of Filipino nationality or another
was em- ployed by the researchers to evaluate the nationality; literate and eager to partici- pate. No payment
respondents’ answers to their proposed plan. Following the or reimbursement of any kind is permitted. The researchers
respondents’ use of the produced application and used social media posts and other recruitment techniques to
completion of the questionnaire, the researchers clarified the urge respondents to participate in the study. Any data that
responses given by the respondents and evaluated the degree the researchers obtained for this study is kept in a safe folder
of agreement on functional adequacy, usability, and security. with a password that only the researchers may access. Only
For the questionnaire to be both ac- curate and effective, the the researchers working on this study have access to
prepared questions must be created precisely using the your data in order to protect your privacy. Online research
investigation came to a conclusion in order to reduce social

IJISRT24JAN1363 www.ijisrt.com 2380

Volume 9, Issue 1, January 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
danger. During a pandemic, researchers use online the problem, they began to allocate roles to everyone for
platforms to direct information to minimize health risks. them to do the responsibilities assigned to them. The duties
The responders are also given instructions and are told that are collecting the data to the client in order to train the
the researchers are happy to offer any clarifications if model or algorithm by the system, creating the user
necessary. To maintain its confidentiality and safeguard interface of the system to interact with the user, developing
your privacy, the data gathered for this study is kept in a the back-end of the system to operate and predict the given
password-protected folder on a private cloud that can only data from the client, and documenting the necessary
be accessed by the researcher. And deletion is the method information that is relative to this study. Every business is
used to discard this material. To keep it private and aware that reviews of their customers are important.
safeguard your identity, personal information like your Therefore, getting the customer’s review is necessary, and
name was substituted with encrypted text, of which only the by doing that, they will know if the business needs to
researchers were aware of the encryption method utilized. improve to meet the satisfaction of their customers. That is
As soon as the research is over, all information about you is why researchers studied and evaluated the client’s data to
destroyed by deleting the data from the encrypted folder. By provide information that may be beneficial for the
signing the Certificate of Consent, you grant the improvement of the company’s product or service while
researchers the right to collect, use, and disseminate the focusing on how to utilize time management in getting the
information you have submitted for this project. Information reviews and processing them to have the analyzed data
that is necessary for this study to be successful. The immediately. By the efficiency of this system, the company
community gains from this research by receiving a system can be informed of the results that they gathered and can
that can aid the companies taking part in it in enhancing improve their services if needed.
their community services. It might be a useful resource for
academics working on related subjects in the future. Table 2 displays the Evaluation Results summary of
the IT Expert in three (3) different criteria assisting the
IV. RESULT AND DISCUSSION potency of the system. The criterion with the highest
mean of 3.59 is the Functional Suitability of the system,
The researchers successfully devised a method for pre- which is equal to Strongly Agree then comes with the
dicting the polarity of sentiment by gathering the client’s Security criterion with a 3.55 mean also having the same
concerns and studying them with each member to deliver interpretation, and lastly the Usability criterion with 3.29
the most efficient method for predicting the polarity of with the interpretation of Agree.
sentiment. After the researchers discussed the solution to

Table 3: Evaluation Result from Experts

Criteria Mean Percentage Interpretation
Functional Suitability 3.59 86.42% Strongly Agree
Usability 3.29 76.54% Agree
Security 3.56 85.19% Strongly Agree
Total 3.48 82.72% Agree

Finally, the overall mean of the criteria is three-point-forty-five3.45 which rated the system as Strongly Agree.

Table 3: Evaluation Result from Endusers

Criteria Mean Percentage Interpretation
Functional Suitability 3.15 71.61% Agree
Usability 3.44 81.48% Agree
Total 3.30 76.55% Agree

Table 3 displays the Evaluation Results summary of V. CONCLUSION

the End- Users in two (2) different criteria assisting the
potency of the system. The criterion with the highest mean This chapter includes a summary of the study
of 3.44 is the Usability of the system, which is equal to ”Sentiment Analysis using Machine Learning: A System to
Agree then comes with the Functional Suitability criterion enhance business process reviews of the Company.”
with a 3.44 mean with also having the same interpretation, Conclusions and suggestions are also included. While
as well as the total mean of both with a 3.29 mean which conducting the study, the researchers were aware of the
evaluated as Agree. significance of sentiment in society, notably in business.
Because they want to know which aspects of their
business need to be addressed, the majority of businesses
are worried about their reputation. The researchers carried
out this experiment to help businesses achieve a more
contemporary method of data analysis because some

IJISRT24JAN1363 www.ijisrt.com 2381

Volume 9, Issue 1, January 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
businesses find it challenging to evaluate customer [3]. Canbolat, Z. and Pinarbasi, F. (2020). Using
feedback. By conducting surveys and interviewing Sentiment Analysis for Evaluating eWOM: A Data
people who were required to participate in the study, the Mining Approach for Marketing Decision Making.
researchers collected the data for this project. By analyzing 10.4018/978-1-52258575-6.ch007.
information using the system’s machine learning algorithm [4]. Henke, N., Bughin, J., Chui, M., Manyika, J.,
or model, the researchers’ newly built system may Saleh, T., Wiseman, B., & Guru, S. (2016). The
forecast the supplied end-user sentiment. To make the Age of Analytics: Competing In A Data-Driven
system work properly, the end user must train a model World. Retrieved from https://www.mckinsey.com/
using a dataset. The accuracy of the model will be /me- dia/mckinsey/business functions/mckinsey
shown if it is complete, allowing the end user to judge analytics/our insights/the age of analytics competing
its suitability. In order to help the end user evaluate the in a data driven world/mgithe-age-of- analyticsfull-
processed data, the system then displays the analyzed report.ashx
data on the dashboard. The researchers came to the [5]. Pons-Muñ oz de Morales, S. (2020). Big data and
conclusion that the developed system could predict the sentiment analysis considering reviews from e-
polarity of the sentiment by having an average commerce platforms to predict consumer behavior.
percentage of 76.55% with a mean value of 3.30, Treballs Finals del Màster de Recerca en Empresa,
interpreted as agreed for the end-users functional, Facultat d’Economia i Empresa, Universitat de
suitability, and usability, while on the other hand, the Barcelona, Curs: 2019-2020. Retrieved from
average percentage given by the IT expert is 82.72% with http://diposit.ub.edu/dspace/handle/2445/173181
a mean value of 3.48. As a result, the system is [6]. Earley, S. and Bernoff, J. (2020). Is Your
currently usable and capable of assisting the client in Data Infrastructure Ready for AI? Harvard Business
data analysis. Future studies should improve the system by Review, 1–5. Retrieved from
automatically converting Tagalog or any dialect to English. https://hbr.org/2020/04/is-your-datainfrastructureready-
This will spare the system’s user the time-consuming task for-ai
of manually deciphering client input. Additionally, as [7]. Pastor, C. (2020). Sentiment Analysis of Filipinos
mentioned by one of the end-users during the study, the and Effects of Extreme Community Quarantine due
system should offer the option to display ratings rather than to Coronavirus (COVID-19) Pan- demic. Poblacion,
polarity, according to the researchers. In order to save Lingayen, Pangasinan, Philippines, 7(7), pp. 91-95.
some time while retaining part of their jobs, the researchers [8]. Okatan, Kağ an. (2020). Machine Learning for
will also suggest this technique to businesses that currently Business Analytics. 10.4018/978-17998-2566-
manually examine customer reviews. In light of this, the 1.ch013.
researchers propose that the developed system be a software [9]. Angenent, Mitch & Pereira Barata, Antó nio &
program that can be accessed from any device, particularly Takes, Frank. (2020). Large-Scale Machine Learning
a mobile one. Additionally, the researchers recommend that for Business Sector Prediction.
the system 10.1145/3341105.3374084.c
[10]. Tyagi, N. (2020). 6 Major Branches of Artificial
Intelligence (AI). Analytics Steps. Retrieved from
ACKNOWLEDGMENT
https://www.analyticssteps.com/blogs/6- major-
branches- artificialintelligence-ai.
The researchers would like to extend their sincere
gratitude and appreciation to everyone who took part in [11]. Bhatt, N. and Swarndeep, S. (2020). Sentiment
Analysis using Machine Learning Technique: A
this study. Their tireless efforts, knowledge, and assistance
were crucial in enabling this investigation. Their continuous Literature Survey. International Research Journal of
support, con- tributions, and cooperation have greatly Engineering and Technology (IRJET), 7(12), pp. 798-
improved this research and helped us accomplish our 802.
objectives. Finally, we would like to extend our sincere [12]. Barba, P. (2019). Sentiment
gratitude to all of the volunteers and study participants Accuracy: Explaining the Baseline and How
without whom the data collection and analysis would not to Test It. Retrieved from
have been possible. We really appreciate your excellent https://www.lexalytics.com/lexablog/sentiment-
contributions to this study. accuracy- baseline- testing.
[13]. Dubovikov, K. (2019). Managing Data Science. Packt
Publishing Ltd. [14] Rokade, P. and Kumari, D.A.
REFERENCES
(2019). Business intelligence analytics using
[1]. Dumbleton, R. (2021). Sentiment analysis: sentiment analysis-a survey. International Journal of
The complete guide to sentiment analysis. Retrieved Electrical and
April 01, 2021, from [14]. Computer Engineering (IJECE). 9. 613.
https://getthematic.com/insights/sentiment-analysis/ 10.11591/ijece.v9i1.pp613-620 [15] Ranjith, P. R.
[2]. Seo, S., Kim, C., Kim, H., Mo, K., Kang, P. (2020). (2019). ML Model Management using Python
Comparative study of deep learning-based sentiment Django Web Framework. In IEEE India Info Vol. 14
classification. IEEE Access, vol. 8, pp.6861-6875, No. 3 (pp. 140-144).
2020, doi: 10.1109/ACCESS.2019.2963426. [15]. Khan M., Malik K. (2019). Sentiment
Classification of Customer’s Reviews About

IJISRT24JAN1363 www.ijisrt.com 2382

Volume 9, Issue 1, January 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Automobiles in Roman Urdu. In: Arai K., Kapoor S.,
Bhatia R. (eds) Advances in Information and
Communication Networks. FICC 2018. Advances in
Intelligent Systems and Computing, vol 887. Springer,
Cham. https://doi.org/10.1007/978- 3-030-03405-4 44
[16]. Saura, J.R., Reyes-Menendez, A., Alvarez-Alonso, C.
(2018). Do online comments affect environmental
management? Identifying factors related to
environmental management and sustainability of hotels.
Sustainability, 10, 3016.
[17]. Kopera, S., Wszendybył-Skulska, E., Cebulak, J.,
Grabowski, S. (2018). Interdisciplinarity in Tech
Startups Development–Case Study of
‘Unistartapp’Project. Found. Manag., 10, 1–10.
[18]. Vanaja, S. and Belwal, M. (2018). Aspect-Level
Sentiment Analysis on ECommerce Data. 2018
International Conference on Inventive Research in
Computing Applications (ICIRCA), 1275- 1279.
[19]. Behdenna, S., Barigou, F., Belalem, G. (2018).
Document Level Senti- ment Analysis: A survey. EAI
Endorsed Transactions on Context-aware Systems and
Applications. 4.154339. 10.4108/eai.14-3-
2018.154339.
[20]. Kabir, A., Karim, R., and Newaz, S. (2018). The
Power of Social Media Analytics: Text Analytics
Based on Sentiment Analysis and Word Clouds on R.
Informatica Economică, p. 25.
[21]. Tribhuvan, Padmapani & Bhirud, Sunil &
Deshmukh, Ratnadeep. (2018). Stacking Ensemble
Model for Polarity Classification in Fea- ture Based
Opinion Mining. Indian Journal of Computer Science
and Engineering. Vol. 9. 91-95.
10.21817/indjcse/2018/v9i3/180903004.
[22]. Singh, J., Singh, G., and Singh, R. (2017).
Optimization of sentiment analysis using machine
learning classifiers. Hum. Cent. Comput. Inf. Sci. 7,
32. https://doi.org/10.1186/s13673-017-0116- 3
[23]. Ziora, L. (2016). The sentiment analysis as a tool of
business analytics in contemporary organizations.
Uniwersytetu Ekonomicznego w Katow- icach,
Zeszyty Naukowe, p. 240.
[24]. Hamdan, H., Bellot, P., Bechet, F. (2016),
Sentiment Analysis in Scholarly Book Reviews,
http://arxiv.org/abs/1603.01595v1 (access: 4.03.2016).
[25]. Liu, G., Xu, X., Deng, B., Chen, S., Li, L. (2016). A
hybrid method for bilingual text sentiment
classification based on deep learning. Proc. 17th
IEEE/ACIS Int. Conf. Softw. Eng. Artif. Intell. Netw.
Parallel/Distrib. Comput. (SNPD), pp. 93-98.
[26]. Huang, Y., Wang, Y., & Ye, F. (2019). A Study of the
application of word cloud visualization in college
english teaching. International Journal of Information
and Education Technology, 9(2), 119-122.

IJISRT24JAN1363 www.ijisrt.com 2383