Volume 1, Issue 1

East Journal of Engineering Publisher: East Publication & Technology

DOI: https://doi.org/
ISSN: 3079-9376

Conversational AI Revolution: A Comparative Review of Machine

Learning Algorithms in Chatbot Evolution
Mohammad Abo Aisha *1 , Rakan Bani Jamei 1
1
Department of Control and Robotics Engineering, Faculty of Engineering and Technology, Al Zaytona University
of Science and Technology, Salfit Street, Al-Laban Al-Sharqiya - Salfit, Palestine.
*Corresponding Author.

Received: 20/01/2025, Revised: 07/02/2025, Accepted: 09/02/2025, Published: 16/02/2025

Abstract
Chatterbots, also known as chatbots, have become essential for improving human-computer interaction in a
number of fields, including e-commerce, healthcare, education, and customer support. From rule-based
systems like ELIZA to contemporary AI-driven solutions employing modern machine learning (ML)
techniques, this review paper examines the development of chatbots. It highlights how ML technologies, such
as decision trees (DT), support vector machines (SVM), linear regression, and natural language processing
(NLP), can be used to build chatbots that are more context-aware, responsive, and adaptive. The paper
highlights important advances including deep learning, multimodal capabilities, and continuous learning
mechanisms by looking at recent advancements and the mathematical models that support these techniques.
These developments have driven an increasing support for chatbots by allowing them to provide personalized
interactions, enhance accessibility, and reduce repetitive tasks. In order to open the door for further study and
applications, this paper aims to bring light on the challenges and the efficacy of using ML into chatbot building.

Keywords: Machine Learning, Chatbots, Decision Tree, Support Vector Machine, Linear Regression,
Natural Language Processing, K-Nearest Neighbor.

1. Introduction
In recent years, Conversational Artificial Intelligence (AI) has emerged as a transformative technology,
revolutionizing human-machine interaction. Chatbots, powered by machine learning (ML) and natural
language processing (NLP), have evolved from simple rule-based systems to sophisticated AI-driven
conversational agents capable of understanding and responding to complex user queries. With
advancements in deep learning, reinforcement learning, and hybrid AI models, chatbots are now widely
deployed across various industries, including customer service, healthcare, education, and e-commerce [1].
This paper provides a comparative review of machine learning algorithms used in chatbot development,
analyzing their evolution, performance, and real-world applications. Traditional methods, such as decision
trees and support vector machines (SVMs), laid the foundation for chatbot intelligence, while modern deep
learning architectures, including recurrent neural networks (RNNs), transformers, and reinforcement
learning models, have significantly enhanced conversational capabilities[2],[3].
The review explores key challenges in chatbot development, including context retention, sentiment
analysis, personalization, and ethical considerations. Additionally, it highlights emerging trends, such as
multi-modal AI chatbots, few-shot learning, and generative AI models like GPT, which are reshaping the
future of conversational AI.
By presenting a comparative analysis of machine learning algorithms, this study aims to provide insights
into the strengths, limitations, and optimal use cases of different chatbot architectures. The findings will

© East Journal of Engineering 1

 Volume 1, Issue 1
East Journal of Engineering Publisher: East Publication & Technology
DOI: https://doi.org/
ISSN: 3079-9376

assist researchers and industry professionals in selecting appropriate ML techniques for developing more
intelligent, efficient, and human-like conversational agents.

1.1 Overview of Chatterbots

Chatbots are now a familiar feature of human-computer interaction in recent years they have become
strategic tools in most organizations. These interactive entities are a new type of software that imitates
human conversation using text or voice to offer various services, from support to learning. AI and NLP
technologies are used to study user inputs to return suitable responses as an organization has the chance to
provide timely services 24/7 with the help of chatbots while augmenting its characteristic processes and
providing the best user experience possible [4].
Chatbots refer to computer programs that mimic a human-like figure with the ability to engage in interaction
and respond according to scripts set or otherwise programmed. It can be integrated with web, mobile, and
other applications like social media messaging services to accommodate different forms. Customer service
interactions, personal assistants, healthcare, online shopping, and education industries apply chatbots most
frequently and in various others. There is a general agreement that the primary reasons for the growing use
of chatbots relate to the fact that they entertain a fast response rate, create as much individual
communication as necessary, and contribute to diminishing human burdens.
The rule-based expert chatbots work through the instructions that are coded into the system and are mostly
inflexible and capable of handling simple tasks and dispelling simple answers. However, NLP and ML
provide more flexible and dynamic responses as compared to a fixed script, all of these developments in
NLP and ML have enabled the correspondingly advanced AI chatbots to be more versatile in their operation
[5].

1.2 History of Chatbots

The history of chatbots started in the 1960s when Joseph Weizenbaum created the first chatbot called
ELIZA at the Massachusetts Institute of Technology. Originally, it was based on simple pattern matching
to imitate conversations, although copying a Rogerian psychotherapist. By today’s standards ELIZA was
primitive, nevertheless, this program proved that machines can indeed have simple conversations with
people [6] [7].
PARRY was created in the 1970’s by the Psychiatrist Kenneth Colby. While ELIZA was programmed
simply to reflect an absence of personality, PARRY was designed to model a patient with schizophrenia
elaborated with a detailed personality as well as language interpretation mechanism [7]. The first-generation
chatbots set a precedent for conversational agents and machines that can engage us in natural language [7].
In the 1990s Richard Wallace created ALICE as an AI program that used the AI Markup Language (AIML)
to improve it are responses [8] [9]. When compared to other models, ALICE was considered an
improvement in chatbot design which even proved capable of winning several AI competitions [9]. Later
in 2001, Smarter Child was able to run on more common messaging services such as AIM and MSN
Messenger as a form of entertainment that included interactive talk [10].
New technologies such as ML and deep learning braced the release of virtual assistants such as Siri by
Apple Inc. (2011), Google Assistant by Google (2016), and Alexa by Amazon (2014) [10]. These systems
are built on top of AI, NLP, and voice recognition to garner contextually intelligent responses, which
created a revolution in digital assistants’ solutions [10]. This evolution has prepared the way for the
conversational models which have been pre-trained on huge datasets, including language text, and
supplemented that with reinforcement learning for great and adaptive responsiveness.

© East Journal of Engineering 2

 Volume 1, Issue 1
East Journal of Engineering Publisher: East Publication & Technology
DOI: https://doi.org/
ISSN: 3079-9376

1.3 Motivation for Researching Chatbots

Chatbots have received a great interest in research and business contexts. The desire to look into chatbots
comes from their possibility to revolutionize user experience and decrease the amount of manual work. By
researching chatbots, we aim to achieve the following objectives:
a. Improve Customer Service: Customer service remains one of the main domains of chatbot applications.
It frees up time because instead of making the user wait for a reply the chatbots can respond to their
queries immediately. Studies on the practical usefulness and performance of these systems, in terms of
user engagement, can improve these systems.
b. Automate Routine Processes: A lot of industries struggle with dealing with recurrent chores and calls.
Chatbots can be used to deal with simple inquiries which means that customer care officers should deal
with more complex problems. Analyzing chatbots allows for understanding the best practices for
integrating them and thus their ability to perform these functions.
c. Enhance Accessibility and Reach: It is important to have a vehicle that isn’t limited in its accessibility
to the average user by things like high costs or even language barriers as those offered by chatbots.
Focused on the discussion of the eccentricity of a certain type of chatbot, the article will help to enhance
the effectiveness of their application for people with different disabilities.
d. Advanced AI and NLP Capabilities: Chatbots are a good example of applying AI and NLP and, at the
same time, they offer a perfect chance to develop these fields. By analyzing the patterns of the
conversation between users and the chatbot and getting feedback from the user, the developers of the
chatbots can enhance the performance of the AI used and how they respond to unfamiliar scenarios.
1.4 Novelty of Modern Chatbots
Current chatbots demonstrate remarkable improvements as compared to the earlier versions, thanks to
improvements in deep learning, NLP, and AI-grounded response generation. Today’s chatbots do not rely
on simple rules hence contrary to traditional chatbots, the new versions involve the use of superior AI
technologies such as that enables them to recognize context and even initiate real-life flows of
conversation and also learn from people’s interactions with them. Key advancements in modern chatbots
include:
a. Deep Learning and NLP: Chatbots present in the current generation use advanced deep learning
models like transformers which help analyze huge amounts of data, learn from context, and produce
sensible output [10].
b. Reinforcement Learning and Human Feedback: Modern chatbots apply reinforcement learning from
people’s feedback to adapt the dialogues they provide [10]. This training approach allows chatbots to
supply likely and contextually proper responses that are also likely to meet the user’s taste and needs.
c. Multimodal Capabilities: Some relativistic chatbot research of the current period has been done on
multimodal chatbots that may recognize and analyze inputs in text, images, and audio. This versatility
enables them to offer more complex user interactions than their fixed-form counterparts which
provides a more realistic feel.
d. Personalization and Context Awareness: AIs in chatbots apply an experience and recall to provide
customized responses to users based on their earlier conversations. This feature improves the
interaction with the users as chatbots can proceed with the recommendations and answers based on
the user input history.
e. Continuous Learning: Unlike primitive, narrow-scope program-based bots, today’s chatbots have
abilities to gain knowledge continuously. In conversations with users, they learn how the users use
language, what new terms are used, and the relevant changing needs of the users; they become more
efficient in the long run.

© East Journal of Engineering 3

 Volume 1, Issue 1
East Journal of Engineering Publisher: East Publication & Technology
DOI: https://doi.org/
ISSN: 3079-9376

2. Machine Learning techniques that are used in chatterbots

This review paper explores at a number of studies that have examined various methods of ML employed in
chatbots.

2.1 Chatbots and Decision Tree

Indriana Widya’s, et al point of view: the increasing use of digital healthcare has led to the development of
maternal and child health mobile health (mHealth) services, mostly through community-based programs
like Posyandu in Indonesia. While health professionals are limited in responding promptly, patients require
media for consultation and decision-making. This study aimed to gather data from midwives and expectant
mothers to create a DT model that would serve as the foundation for a semi-automated chatbot. Pregnant
women (n = 10) and midwives (n = 12) participated in semi-structured focus group discussions (FGD) in
March 2022 using an exploratory qualitative methodology. Three main topics emerged from the results,
which included 38 codes, 15 categories, and 7 subthemes: health monitoring, information on maternal
health services, and maternal health education. To ensure quality, the DT technique was applied to these
themes based on user needs, evidence, and expert sources. In conclusion, maternal health and monitoring
education can benefit from the use of a semi-automated chatbot, where serious situations should be
communicated with midwives in a non-automated manner. Using the DT method helped with early
identification, supported a clinical choice, and guaranteed high-quality content. Additionally, user
evaluation must be measured in future studies [11].

According to Dillon Chrime, COVID-19 has affected healthcare systems and billions of people worldwide.
However, as of right now, there isn't an open-source chatbot that can help patients and important healthcare
stakeholders assess the possible severity of a COVID-19 infection or recognize other illnesses and bodily
reactions that could increase the risk of developing severe cases of the virus. This initial study investigates
the deficiency of a DT with binary classification of cases and non-cases based on age, bodily system, viral
infection, comorbidities, and any symptoms in the development of the COVID-19 "case by case" chatbot.
A DT that explained the tools required for stratification for the chatbot application and for interacting with
the user using a set of relevant nodes was created after the literature review. 212 nodes were constructed
from the total number of lucid nodes, which were then categorized by body systems, medical conditions,
comorbidities, and relevant manifestations seen in the literature, ranging from heart diseases to lung
disorders. This produced 63560 "what if" choices, explained why severe COVID-19 cases can be difficult
to comprehend in some situations, and provided a way to understand the data required to validate the DT.
The DT illustrates how process strengthening is achieved by classifying the viral infection by bodily system
and by include comorbidities and symptoms in the framework. This prototype application offers insight
into the kind of data needed for decision support help, even though the practical clinical DT is not suitable
for COVID-19 cases [12].

Ibrahim Alnedhami’s, et al point of view: As technology advances in the modern era, AI has become
increasingly popular across all relevant domains. Indeed, automated chat systems, sometimes known as
"smart conversation" chatbots, are one of these crucial areas. In this work, we present a smartphone app
that allows pregnant women to chat automatically. To provide a healthcare assistant, they have utilized this
technology. Every mother wants to bring a child into the world, thus the suggested health care system assists
expectant mothers by talking with them about certain symptoms, identifying illnesses, and forecasting the
delivery method. It serves a variety of purposes for expectant mothers, including advising them on safe
medicine, healthful activity, and nutrition. The suggested chatbot system was developed using the ML
approach, namely the DT model, rather than relying solely on rules. With the aid of ML techniques, user
inputs that were not previously entered during training can be processed and answered with a high degree
of accuracy. In addition, the system supports words with diacritical marks, fixes the majority of spelling
and grammar mistakes, and has many other capabilities. The smart adviser was evaluated using a variety

© East Journal of Engineering 4

 Volume 1, Issue 1
East Journal of Engineering Publisher: East Publication & Technology
DOI: https://doi.org/
ISSN: 3079-9376

of conversational formulations after it had been trained multiple times. The accuracy of the evaluation was
85.45% [13].

According to Pavel Smutny et al.: In particular, the usage of chatbots as a way to communicate with the
target audience has grown in popularity over the last ten years due to the widespread adoption of mobile
devices. As a result, the use of these tools is expanding at a tremendous rate. In many contexts, these
portable devices transform communication and promote learning in a variety of ways. This study
concentrated on the usage of Facebook Messenger teaching chatbots to increase comprehension. The
independent internet directory was examined in order to evaluate chatbots for this study, and 89 distinct
chatbots were found. The author categorized each of the chatbots found in the study according to the type
of development platform, language feature, and discussion topic. Lastly, 47 instructional chatbots on
Facebook Messenger were assessed using the analytical hierarchy method in terms of teaching, impact,
accessibility, and humanity. Our study demonstrated that using Facebook Messenger chatbots for education
is as easy as suggesting educational topics and sending tailored messages. According to the findings,
chatbots that are now being developed as part of instant messaging apps are still in their infancy as a
teaching tool for AI. The outcome gives teachers a lesson to help them understand how to use chatbots in
the classroom and a variety of chatbots to test out as recommended [14].

2.2 Chatbots and Support Vector Machine

Viewpoint of Show-Jane Yen et al. The issue of choosing and organizing an increasing volume of text data,
to which access is frequently essential for information technology and Internet services, is a surprisingly
pertinent problem. Over the past ten years, ML solutions—such as AI and pattern recognition—have proven
to be highly effective and adaptable in a variety of industries. The technique of looking through large
document collections for precise answer sentences is known as question answering (QA). A question
classifier, simple document/passage retrievers, and the proposed context-ranking models are all parts of the
ML based question answering system that this paper claims to have. The context-ranking model is
instructed to reorder the passages that the initial retrievers found based on the question classifier's answer
type. Learners can use this method to achieve flexible instructional characteristics, including word forms,
syntactic aspects, and semantic word aspects. The proposed context-ranking model takes into account
whether or not the input passage is pertinent to the query type while annexing rich information. The
successive labeling of tasks is the foundation for that. We use question categorization standards and TREC-
QA tracks to assess the suggested methodology. According to the accuracy test, the question classifier's
accuracy was 85.60% when no further semantic or syntactic taggers were introduced. This accuracy
increased to 88.60% when the extended term approach methods suggested in this paper and a fixed related-
word list were included. The QA model using the gold TREC-provided relevant document set achieves an
MRR rank of 0.563 on the TREC-10 QA test. The mean reciprocal rank for retrieving simple documents
and passages is 0.342 [15].

The perspective of Jianwei Gong and others: When discussing driver assistance systems, and autonomous
vehicles specifically, the issue of road detection is quite important. This research focuses on feature
extraction and classification for front-view road identification. More precisely, we suggest a quick method
for road detection using SVM and live self-supervised learning. The use of online training into the suggested
road identification algorithm improves the system's flexibility and lowers the frequency of incorrectly
identifying roads and non-road classifications by updating the training data. The strategy shown here can
also be seen as another paradigm for self-supervised online learning in the context of using a classification-
based approach for intelligent vehicle road identification [16].

David Agustriawan’s, et al point of view: Every industry has seen a sharp rise in technological progress,
particularly the healthcare sector. During this time, hospital management began to advance through the use
of technical instruments and systems. Patient data can be saved and prepared methodically to be utilized as
an appointment queue line once the hospital has configured the system. A chatbot that uses NLP to boost

© East Journal of Engineering 5

 Volume 1, Issue 1
East Journal of Engineering Publisher: East Publication & Technology
DOI: https://doi.org/
ISSN: 3079-9376

the effectiveness of healthcare services could be used to improve it. An ideal classifier that learns the
classification hyperplane in a space map with the maximum distance (margin) to the training samples is the
SVM approach. The SVM will forecast the recommended specialist based on the users' comorbidities and
symptoms [17].

Melanie Lourens’s, et al point of view: Appropriate healthcare is essential to a healthy lifestyle.

Nevertheless, making an appointment with a doctor for any health condition might be quite difficult. The
objective is to create a ML based medical chatbot that can recognize the ailment and provide basic
information about it before a patient contacts a doctor. Using a medical robot will improve access to medical
information and reduce healthcare costs. The bots are computer models that use speech recognition to
interact with humans. The chatbot stores the information in a database so that it can identify the language
phrases, choose a query, and respond. Text similarity and ranking are computed using mutual information,
TF-IDF, and N-gram. For the given query, more similar phrases will be located, and each verb in the input
word will be given a score. A supplier, an expert piece of software, answers questions that are ambiguous
or not present in the system [18].

2.3 Chatbots and linear regression

Arunee Ratikan’s, et al point of view: These days, daily routines include consuming junk food, not having
enough time for exercise, and other things that cause health problems for older adults. The majority of
senior citizens frequently experience high blood pressure as they age. Numerous serious illnesses can result
from this condition. Furthermore, there are not many doctors working in the hospital. To provide prompt
care, the doctor typically takes a brief note about symptoms, which may not be sufficient for a diagnosis.
As a result, we suggested a chatbot for daily health monitoring for senior citizens. To establish a personal
health record (PHR), we must gather information from the elderly. We create conversational chatbots to
communicate with senior citizens through the LINE app. The results of this study complement the doctor's
work because the doctor is better able to diagnose illnesses and provide treatment recommendations after
reading the daily PHR. Additionally, the linear regression technique was created to track the trend of older
people's blood pressure. As a result, they can prevent or alleviate certain illnesses through chatbot alerts
and health maintenance [19].

R Thamilselvan’s, et al point of view: Higher education institutions are incorporating AI to improve their
website support and user experience in light of the upcoming technology revolutions. A chatbot built on
GPT-2 has been created to help stakeholders, instructors, and students navigate college websites with ease.
This research study's goal is to assist college websites by responding to user inquiries and forecasting
students' cut-off marks depending on their academic achievement. To answer questions on admissions,
course information, academic resources, administrative processes, and campus life, this AI-powered
chatbot uses ML and NLP techniques. The developed chatbot, which focuses on Q&A exchanges, uses
deep learning techniques to provide thorough and contextually relevant responses. The chatbot, which is
based on OpenAI's GPT-2 model, is adept at producing text that appears human and is used in a variety of
settings, such as information retrieval and customer service. Specifically, it uses regression models
including Linear Regression, DT Regression, SVM Regression, Random Forest Regression, and XGBoost
Regression to estimate admission cut-off marks. These models take into account several variables, including
academic performance, department preferences, board exam difficulty, prior cut-off lists, joining year,
quotas, and employment market demands. The prediction accuracy of the suggested chatbot is evaluated
using performance evaluation metrics such as Mean Squared Error (MSE), Root Mean Square Error
(RMSE), and R-squared. The chatbot improves the entire user experience of educational websites by
thoroughly analyzing past student data to provide users with full replies and precise predictions about future
academic success [20].

According to Cecilie Bertinussen Nordheim et al., chatbots are anticipated to play a significant role in
customer service. The degree to which users trust these chatbots also affects how often they are used.

© East Journal of Engineering 6

 Volume 1, Issue 1
East Journal of Engineering Publisher: East Publication & Technology
DOI: https://doi.org/
ISSN: 3079-9376

However, the level of trust that people have in chatbots is still unknown. We propose a questionnaire survey
with 154 participants that examines factors for customer service chatbot trust in order to close this
knowledge gap. There were two sections to the study: An interpretive review of the significance of the well-
established antecedents of trust in interactive systems, as well as a preliminary examination of other relevant
factors for chatbot trust. In order to achieve this, one of four customer support chatbots was used to convert
the participants. In light of the study's findings, we propose a preliminary research model of confidence in
service chatbots that takes into account the inclination for technology adoption, perceived risk and brand,
and perceived chatbot attributes (expertise and reaction time) [21].

2.4 Chatbots and natural language processing (NLP)

According to Suman Rajest et al., the goal of this project is to create a chatbot that will employ NLP to help
its users. Chatbots are AI tools that conduct user interviews, comprehend user inquiries, provide answers,
and help users in whatever manner they can. The goal of this application is to create a chatbot for customer
service that can interact with customers and offer assistance in a few specific situations. This chatbot will
process messages containing basic user requests, automatically assign a tag from one of the pre-existing
tags, and provide the appropriate response. If the bot determines that the questions being asked are
extremely complicated, it will forward the conversation to a real human assistant. The Natural Language
Tool Kit (NLTK) and PyTorch (a Python Deep Learning package) will be used to develop the ML model
upon which the Chatbot will be based. The model being used here is the feed forward neural network. The
input layer, hidden layers, and output layer are the three layers that make up this neural network. The total
number of unique words in the data set is equally distributed between the input and hidden layers of the
network. Conversely, the number of nodes in the output corresponds to the number of distinct tags into
which the provided data set is divided. This kind of neural network is ideal for the basic development of
chatbots because it can be trained or used with little computational resources. Our coffee shop design has
the ability to order coffee, make jokes, recommend drinks, and more. Because of this chatbot's rather high
degree of customisation, many alternatives can be implemented into various circumstances. One significant
feature of this chatbot is that, although the neural network itself seldom needs to be changed, the dataset it
is trained on may be readily modified to include new tags. Because of this, the model is quite reliable.
Chatbots like this one are being used in a variety of businesses, including government agencies, automated
food delivery services, and the medical sector. Additionally, this feature makes it very simple to alter the
chatbot's training and testing [22].

Lalwani et al.'s perspective: Software applications offer a variety of user interfaces, such as menus, forms,
graphical user interfaces (GUIs), command lines, and natural language. Although UI can be divided into
web-based and GUI types, there may be situations in which a different type of UI is required. This is why
we require an interactive interface that allows users to communicate with information retrieval systems like
chatterbots. The chatbot is a type of bot that has been accessible on chat networks. This is the tendency that
has been noticed: the user can invoke them through graphical user interfaces or widgets. They frequently
offer a stateful service, meaning that the application stores the data from each session. On a college's
website, it's frequently impossible to even determine where to seek for information. It becomes challenging
to obtain information when one does not work at or attend a specific university or college. A college inquiry
chatbot provides a solution to the problems; it is a standard and educational widget that is quick, easy, and
educational. Enhancing college websites' homepages gives users useful information that improves their
experience. These days, chatbots—intelligent systems—are created using AI and NLP techniques. This
interface manages a number of tasks, including exam cell, admission, academics, user attendance, grade
point average, placement cell, and many more [23].

The viewpoint of Mustafa Raza Rabbani et al. AI is currently being seen to improve the banking and finance
sector. It has undergone significant development and is probably going to continue to be a popular
advancement for some time to come. The current study examines the impact that technologies play in the

© East Journal of Engineering 7

 Volume 1, Issue 1
East Journal of Engineering Publisher: East Publication & Technology
DOI: https://doi.org/
ISSN: 3079-9376

banking and finance industry as well as whether the use of AI has altered how banks and other financial
services providers operate. One of the most important aspects of the banking and financial sector is
customer interaction. For customers of Islamic banking and finance, this study suggests a chatbot platform
based on NLP and AI. The suggested chatbot is currently the first of its kind to assist Islamic banking and
finance till clients receive interactive Islamic financial solutions that comply with shariah for their particular
needs [24].

Mohammed Benhmed and associates' perspective: Chatbots, also known as conversational agents, are
conversational software programs that can communicate with actual people using a natural language
interface. One of the fundamental tasks in AI and NLP is dialogue management. The best talking chatbot
has taken years to construct, according to AI leaders. Although chatbots are capable of a wide range of
functions, their primary function is to comprehend spoken language and initiate a suitable response. In the
past, chatbot architectures were created using handwritten code, pre-established templates and conventions,
or basic statistical techniques. Around 2015, end-to-end neural networks took the role of these models due
to their improved learning capabilities. The encoder-decoder recurrent model is currently the modeling
approach of conversations that is used more commonly. The scientific field of neural machine translation
is where this approach was applied, and it shows impressive performance. To improve the conversational
capabilities of chatbots, numerous features and modifications have been implemented thus far [25].

According to Hussam Abdulla et al., viewers are moving toward a society that will welcome more
interaction between humans and machines as a result of AI’s development in numerous economic sectors.
Because of this, there are subfields like NLP that are used in chatbots and other technologies. In this regard,
the current study uses recent and critical secondary research to discuss the development of chatbots using
NLP by embracing AI and machine linguistics. These technologies allow devices to analyze text and
facilitate text planning, sentence planning, and text realization for NLP [26].

The perspective of Vitaliy Husak and associates: The work focuses on developing an information system
that uses chatbots and NLP to provide a list of stylish clothing recommendations that satisfy users' needs
or expectations. It offers both scholarly and real-world applications for chatbots as voice interfaces for NLP.
The goal of this project is to create software that would enable chatbot functionality in the already-available
teleGRAM messenger. When installing other chatbots, similar chatbots with comparable functionality were
examined, along with their benefits and drawbacks. A portion of the system's server is constructed with
Java and the Spring Framework to process incoming data. IntelliJ IDEA is a programming environment in
the form of an Integrated programming Environment, or IDE for short. In addition to attaining a high level
of user language recognition, which has encouraged interaction and communication, NLP was also
incorporated into the system through the DialogFlow service. The Bot API is used to establish a connection
with the Telegram messenger. Once more, Hibernate is used as the data ORM for processing our database,
while Jsoup is used for our parsers or for extracting web content [27].

2.5 Chatbots And K-Nearest Neighbour

Viny Christanti’s, et al point of view: Nowadays, everything is done online, including office work, teaching,
and learning, particularly in the domain of school administration. An informant must be able to meet the
academic and non-academic information demands of students' parents and other stakeholders during the
activity. One way to get around these requirements is with chatbots, which offer features that let users ask
questions about where and when. Thus, it can be concluded that chatbots can be a public tool, particularly
for schools, which can use them as a source of school information outside of their operations. The Modified
K-Nearest Neighbor (MKNN) approach was used to create this chatbot. The most recent iteration of the K-
Nearest Neighbor approach is called MKNN, and in its application, it performs additional steps after
calculating Euclidean Distance and weight voting. In order to determine the shortest distance to the data
class that is being addressed by carrying the weight of information biggest or most, the chatbot will first
determine the value of the shortest distance to the database. Once this is done, it will calculate the value

© East Journal of Engineering 8

 Volume 1, Issue 1
East Journal of Engineering Publisher: East Publication & Technology
DOI: https://doi.org/
ISSN: 3079-9376

weight or information from that distance. This is the premise of the weight voting method itself. This makes
it easier for the deep chatbot to identify a database's category and deliver more precise and certain responses
[29].

From the perspective of Mr. PVM Vijay et al. Chatbots are gradually taking the place of humans in jobs
that were formerly exclusively performed by humans, such as customer service or teaching. From basic
rule-based chatbots to the current advanced AI chatbot era, the role of chatbots has expanded. These days,
chatbots can effortlessly carry on a conversation by customizing their previous interactions with customers.
This project's objective is to create a conversational robot that can be integrated into OpenAI's GPT-Chat.
This novel approach has the potential to refocus future chatbot research in a particular direction. Based on
the findings of our study, we propose a number of research avenues during our examination [29].

According to Omar Khadeer Hussain et al., we are living in the age of intelligent machines. As demonstrated
by AI, ML, and deep learning, they are now also learning, appearing, and occasionally sobbing like humans.
A chatbot, which is a conversational software agent activated using NLP, is a good illustration of such a
computer. An overview of existing chatbots and the methods used in them is given in this study. The
evolution, distinctions, and comparable features of the existing chatbots are specifically covered. We
examined the features and technical details of 11 top chatbot application platforms by analyzing their
attributes. According to a report, about 75% of customers are dissatisfied with the customer service they
receive, and it might be difficult to come up with informed, thorough, and time-consuming responses these
days. Historically, chatbot development techniques have depended on manually generated templates and
rules. It was not long until end-to-end neural networks effectively replaced traditional models, and deep
learning was just around the corner. To put it another way, Deep Neural Networks are the most effective
generative-based model for solving conversational response generation issues. The authors of this study
provisionally reviewed over 70 publications on the subject of chatbots that were published in the last five
years based on a review of the literature. Overall, based on the literature review, this study compared a few
selected studies using the technique employed. Additionally, this study explained how existing chatbot
models affect conversation quality since they ignore certain aspects when producing responses [30].

According to Mutiwokuziva et al., in order to accomplish this, we constructed a chatbot in this study to
evaluate the feasibility of teaching computers to translate natural language writing in the form of free-text
and text messages. Using firsthand experience, this paper attempts to comprehend the methods, skill, and
power of NLP with recurrent neural networks (RNN). We began our experiment using Google Word2vec
and a sequence-to-sequence LSTMs cell neural network in order to do this. As a result, our results show
that the type of output quality prediction is dependent on the number of training sessions and the caliber of
the language model that was used to train our model bot. In this instance, an RNN-based chatbot
demonstrates both generative abilities and thinking [31].

From the perspective of Akhtar Rasool et al: A conversational agent, commonly known as a chatbot, is a
computer program that mimics human communication abilities. The development of communication is one
of the most important phases in the creation of any type of chatbot. Even though AI and NLP have advanced
significantly, developing a solid chatbot model remains a significant issue. A conversational bot can
accomplish a lot of things. Stated differently, they ought to provide pertinent responses and have a basic
understanding of the user's objective. Through the use of a vocal interface, this software program enables
users to communicate with one another just as frequently as they would with a human. Because of this, they
are utilized in practically every setting for client communication, including social media. These days, two
basic models are used when developing a chatbot. two kinds of models: generative and retrieval-based.
Thus, the earlier methods that relied on statistical techniques or handwritten instructions and patterns have
been superseded by the breakthroughs in new approaches to AI and deep learning, such as end-to-end
trainable neural networks. The current study suggests a different kind of deep learning, namely the use of
deep neural learning to build a chatbot. When processing input and using the network's output, this method
uses a multi-layer neural network [32].

© East Journal of Engineering 9

 Volume 1, Issue 1
East Journal of Engineering Publisher: East Publication & Technology
DOI: https://doi.org/
ISSN: 3079-9376

According to Bhartiya et al., chatbots have proven to be successful in lowering people's workloads since
they offer automated solutions that mimic human involvement in a variety of business and societal
challenges. The design and implementation of a university counseling auto-reply bot to respond to inquiries
regarding engineering at the university level are thoroughly described in this work. The problem of
overfitting was resolved once we applied the appropriate NLP to the JSON-formatted data from our
university, which had been trained using a feedforward neural network. The user will be able to view the
message in the Facebook Messenger interface as a useful tool for communication the next time they log in
after the chat application has been linked into Facebook Messenger. The probability scores of correct
answers increased somewhat from 0.46 for the first end-user test phase to 0.72 for the second end-user test
phase when additional training phrases and keywords were added to the dataset [33].

3. Mathematical models of ML algorithms

3.1 Decision Tree’s Mathematical Model
A DT consists of decision points represented by nodes (N), where edges (E) indicate the outcomes of splits
or tests performed at these nodes, and leaves (L) serve as terminal nodes that provide either a value for
regression or a class label for classification. When building a tree, it is essential to split a dataset S at each
node. The working principle of DT algorithm as shown in Figure 1. Typical standards consist of:
a. Gini Impurity Classification

G(s) = 1 - ∑𝑘𝑖=1 𝑝𝑖 2 (1)

Where pi is the proportion of samples in S belonging to class i, and k is the number of classes [35].
b. Information Gain Classification
|𝑆𝑣|
IG (S, A) = H(S) – ∑𝑣∈𝒗𝒂𝒍𝒖𝒆𝒔(𝑨) |𝑆|
𝐻(𝑆𝑣) (2)

Where:
H(s) = - ∑𝑘𝑖=1 𝑝𝑖 𝑙𝑜𝑔2 (𝑝𝑖) (3)
is the entropy of the subset S, and Sv is the subset of samples for which feature A has value v [36].
c. Variance Reduction Regression
|𝑆𝑣|
V R (S, A) = Var(𝑆) – ∑𝑣∈values(A) |𝑆|
Var(𝑆𝑣) (4)

Where:
1
Var(𝑆) = |𝑆|
− ∑𝑖∈S(𝑦𝑖 − 𝑦̅)2 . (5)

Here, 𝑦𝑖 is the target value, and 𝑦̅ is the mean of 𝑦 in S [34].

d. Conditions of Stopping
When one or more of the following circumstances are satisfied, the recursive splitting comes to an end, the
maximum depth, d max, is attained. A node's minimum sample count (𝑛min n min) is less than a threshold.
Impurity and variance are not considerably reduced by additional splits [38]. Prediction for a new input x:
from the root to a leaf on the tree. To follow the associated edge, assess the circumstances at each node. At
the leaf, return the value:

© East Journal of Engineering 10

 Volume 1, Issue 1
East Journal of Engineering Publisher: East Publication & Technology
DOI: https://doi.org/
ISSN: 3079-9376

• Classification: The leaf's majority class label [39].

• Regression: The leaf's mean of the target values [39].
In terms of mathematics:
𝑐𝑖 𝑖𝑓 𝑥 ∈ Lead 𝑖 with class label 𝑐𝑖
𝑇(𝑥) = { (6)
𝑦̅𝑖 𝑖𝑓 𝑥 ∈ Lead 𝑖 with mean target 𝑦̅𝑖

Figure 1: DT working principle flowchart [34].

3.2 Support vector machine’s mathematical model
Finding the hyperplane that best divides the data points from several classes is the main concept. The
working principle of SVM algorithm as shown in figure 2.

a. Linear classification
Given set of training data {(𝑥𝑖, 𝑦𝑖)} where 𝑥𝑖 ∈ ℝ𝒏 (input vectors) and 𝑦𝑖 ∈ {-1, +1} (class labels), the
SVM seeks a hyperplane 𝑤. 𝑥 + 𝑏 = 0 that maximizes the margin between the two classes. The objectives
are to find the optimal values of 𝑤 and 𝑏 that satisfy the following constraints:
𝑦𝑖(𝑤. 𝑥𝑖 + 𝑏) ≥ 1, ∀𝑖 = 1, 2, … , 𝑁 (7)
Where:
𝒘 is the weight vector normal to the hyperplane.
𝒃 is the bias term, shifting the hyperplane.
𝑵 is the total number of training samples [41].

© East Journal of Engineering 11

 Volume 1, Issue 1
East Journal of Engineering Publisher: East Publication & Technology
DOI: https://doi.org/
ISSN: 3079-9376

b. Maximizing the Margin

The margin between the two classes is defined as the distance from the closest point to the hyperplane, and
the goal is to maximize this margin. The margin is given by:
1
Margin = (8)
||𝑤||

1 1 2
Thus, the objective becomes to maximize , which is equivalent to minimizing 2
||𝑤|| . The
||𝑤||
optimization problem becomes:
𝒎𝒊𝒏 𝟏 𝟐
(𝒘,𝒃) 𝟐 ||𝒘||

subject to
𝑦𝑖(𝑤. 𝑥𝑖 + 𝑏) ≥ 1,
∀𝑖 = 1, 2, … , 𝑁 [41].
c. Kernel Trick (Nonlinear Classification)
For linearly non-separable data, SVM uses a kernel function 𝐾(𝑥, 𝑥́ ) to map the input data into a higher-
dimensional space where a linear separator can be found. The decision rule becomes:
𝑓(𝑥) = 𝑠𝑖𝑔𝑛( ∑𝑁
𝑖=1 𝑎𝑖 𝑦𝑖 𝐾 (𝑥, 𝑥́ ) + 𝑏) (9)
Where:
𝒂𝒊 are Lagrange multipliers determined by solving the optimization problem.
𝑲(𝒙, 𝒙́ )is the kernel function, commonly the Radial Basis Function (RBF), polynomial kernel, etc [42].
d. Dual Problem
To solve the optimization problem, the Lagrangian dual form is often used. The dual formulation of the
SVM optimization problem is:
𝑚𝑎𝑥 ∑𝑁 1
𝑎 ( 𝑖=1 𝑎𝑖 − ∑𝑁
𝑖𝑗=1 𝑎𝑖 𝑎𝑗 𝑦𝑖 𝑦𝑗 𝐾(𝑥𝑖, 𝑥𝑗)) (10)
2

Subject to the constrains:

0 ≤ 𝑎𝑖 ≤ 𝐶, ∑𝑁
𝑖=1 𝑎𝑖 𝑦𝑖 = 0 (11)
Where 𝐶 is the regularization parameter that controls the trade-off between maximizing the margin and
minimizing the classification error [43].

© East Journal of Engineering 12

 Volume 1, Issue 1
East Journal of Engineering Publisher: East Publication & Technology
DOI: https://doi.org/
ISSN: 3079-9376

Figure 2: SVM working principle flowchart [40].

3.3 Natural language processing’s mathematical model
A branch of AI called NLP studies how computers and human languages interact. Enabling machines to
comprehend, interpret, and produce human language is the aim. Numerous mathematical models and
methods that focus on language's structure, grammar, and semantics are used in this process [44]. Figure
(3) shows a NLP algorithm used to build a healthcare chatbot model.

© East Journal of Engineering 13

 Volume 1, Issue 1
East Journal of Engineering Publisher: East Publication & Technology
DOI: https://doi.org/
ISSN: 3079-9376

Figure 3: NLP chatbot model [45].

a. Tokenization
Tokenization is one of the first steps in NLP, where text is split into smaller units, such as words or
sentences, called tokens. Mathematically, tokenization can be represented as a function 𝑇: ℒ → ℒ ∗, where
ℒ represents the input language (sequence of characters) and ℒ ∗ represents the output sequence of token
[46].
b. Part-of-Speech (POS) Tagging
In POS tagging, each token in a sentence is assigned agrammatical label, such as noun, verb, or adjective.
The process can be mathematically expressed as a function 𝑃: ℒ ∗ → ℙ, where ℒ ∗ is the set of tokens and ℙ
is the set of part-of-speech labels. Probabilistic models, such as Hidden Markov Models (HMM), are
commonly used to model this task [47].
c. Named Entity Recognition (NER)
NER involves identifying entities such as names, locations, and organizations from text. This can be framed
as a sequence labeling problem. Mathematically, for a sequence of tokens 𝑥1, 𝑥2, … … … , 𝑥𝑛, NER assigns
each token a label from
a set 𝐿 = {𝑃𝑒𝑟𝑠𝑜𝑛, 𝐿𝑜𝑐𝑎𝑡𝑖𝑜𝑛, 𝑂𝑟𝑔𝑎𝑛𝑖𝑧𝑎𝑡𝑖𝑜𝑛, … … . . }, 𝑠𝑢𝑐ℎ 𝑡ℎ𝑎𝑡:
𝑦𝑖 = 𝑓(𝑥𝑖), ∀𝑖 = 1, 2, … , 𝑛 (12)
Where 𝑦𝑖 represents the entity label assigned to each token 𝑥𝑖 [48].
d. Sentiment Analysis

© East Journal of Engineering 14

 Volume 1, Issue 1
East Journal of Engineering Publisher: East Publication & Technology
DOI: https://doi.org/
ISSN: 3079-9376

Sentiment analysis assigns a sentiment label (positive, negative, or neutral) to a given piece of text.
This can be represented as:
𝑆 = 𝑓(𝑇 ) (13)
Where 𝑆 is the sentiment label and 𝑇 is the tokenized input text. ML models like SVM or deep learning
architectures such as RNN[49] are often used for sentiment classification [50].
e. Word Embeddings

Word embeddings are vector representations of words that capture semantic relationships between words.
One common approach to word embeddings is through techniques. The embedding for a word 𝑤 is
represented as a dense vector 𝑣𝑤 ∈ ℝ𝒅 , where 𝑑 is the embedding dimension. The relationship between
words can be captured by the cosine similarity between their vectors:
𝑣𝑤 .𝑣𝑢
𝐶𝑜𝑠(⍬) = (14)
||𝑣𝑤 || .||𝑣𝑢 ||

Where 𝑣𝑤 and 𝑣𝑢 are the word vectors for words 𝑤 and 𝑢, and ⍬ is the angle between them [51].

f. Machine translator

Machine translator involves translating text from one language to another. In statistical machine translation
(SMT), this process can be represented using a probability distribution:

𝑃(𝑠|𝑡) 𝑃(𝑡)
𝑃(𝑡|𝑠) = 𝑃(𝑠)
(15)

Where 𝑃(𝑡|𝑠) is the probability of target sentence 𝑡 given the source sentence 𝑠, 𝑃(𝑠|𝑡) is the translation
model, 𝑃(𝑡) is the language model for the target sentence, and 𝑃(𝑠) is the language model for the source
sentence [48].

4. Conclusion
The integration of ML algorithms has significantly enhanced the capabilities of chatbots, transforming them
from simple, rule-based systems into sophisticated tools capable of understanding and generating human-
like responses. Advanced techniques, such as DT, SVM, linear regression, and NLP, enable chatbots to
provide personalized, context-aware interactions across various domains, including healthcare, customer
service, and education. Despite the remarkable progress, challenges remain in areas like accuracy,
adaptability, and trustworthiness, emphasizing the need for ongoing research to further improve the
functionality and reliability of these intelligent conversational agents.

References
[1] D. Salman, N. A. Musalhi, M. Kusaf and E. Celebi, "Integration of Electric Vehicle Charging Stations
into the Unit Commitment Modeling," 2022 11th International Conference on Power Science and
Engineering (ICPSE), Eskisehir, Turkey, 2022, pp. 79-83

[2] Diaa Salman, Abdirahman Farah Ali, Suleiman Abdullahi Ali, Abdullahi Sheikh Mohamed, "Enhancing
Power Grid Stability through Reactive Power Demand Forecasting Using Deep Learning," SSRG
International Journal of Electrical and Electronics Engineering, vol. 11, no. 12, pp. 170-185, 2024.

© East Journal of Engineering 15

 Volume 1, Issue 1
East Journal of Engineering Publisher: East Publication & Technology
DOI: https://doi.org/
ISSN: 3079-9376

[3] Diaa Salman, Cem Direkoglu, Nabeel Altanneh, Abdulaziz Ahmed, "Hybrid Wavelet-LSTM-
Transformer Model for Fault Forecasting in Power Grids," SSRG International Journal of Electrical and
Electronics Engineering, vol. 11, no. 12, pp. 314-326, 2024.
[4] Vashishth, Tarun Kumar, et al. "Artificial Intelligence (AI)–Powered Chatbots: Providing Instant
Support and Personalized Recommendations to Guests 24/7." Technology and Luxury Hospitality.
Routledge, 2024. 211-236.
[5] Bataineh, Abdallah Q., et al. "Enhancing natural language processing with machine learning for
conversational AI." IET Conference Proceedings CP870. Vol. 2023. No. 39. Stevenage, UK: The Institution
of Engineering and Technology, 2023.
[6] Natale, Simone. "If the software is narrative: Joseph Weizenbaum, artificial intelligence and the
biographies of ELIZA." New Media & Society 21.3 (2019): 712-728.
[7] Zemčík, MgrT. "A brief history of chatbots." DEStech Transactions on Computer Science and
Engineering 10 (2019).
[8] AbuShawar, Bayan, and Eric Atwell. "ALICE chatbot: Trials and outputs." Computación y
sistemas 19.4 (2015): 625-632.
[9] Bani, Balbir Singh, and Ajay Pratap Singh. "College enquiry Chatbot using ALICE." International
Journal of New Technology and Research 3.1 (2017): 263368.
[10] Soofastaei, Ali. "Introductory chapter: Virtual assistants." Virtual Assistant 1 (2021).
[11] Puspitasari, Indriana Widya, et al. "Development of a Chatbot for Pregnant Women on a Posyandu
Application in Indonesia: From qualitative approach to decision tree method." Informatics. Vol. 9. No. 4.
MDPI, 2022.
[12] Chrimes, Dillon. "Using decision trees as an expert system for clinical decision support for COVID-
19." Interactive Journal of Medical Research 12.1 (2023): e42540.
[13] Mokhtar, Roqaiah, et al. "A Smart Advisor for Pregnancy Healthcare Using Chatbot Approach Based
on Decision Tree Methods." 2024 4th International Conference on Emerging Smart Technologies and
Applications (Marta). IEEE, 2024.
[14] Smutny, Pavel, and Petra Schreiberova. "Chatbots for learning: A review of educational chatbots for
the Facebook Messenger." Computers & Education 151 (2020): 103862.
[15] Yen, Show-Jane, et al. "A support vector machine-based context-ranking model for question
answering." Information Sciences 224 (2013): 77-87.
[16] Zhou, Shengyan, et al. "Road detection using support vector machine based on online learning and
evaluation." 2010 IEEE Intelligent Vehicles Symposium. IEEE, 2010.
[17] Matthew, Ryan, et al. "The Development of A Medical Chatbot Using The SVM Algorithm." 2022
4th International Conference on Cybernetics and Intelligent Systems (ICORIS). IEEE, 2022.
[18] Panduro-Ramirez, Jeidy, and Durga Prasad Gangodkar. "Artificial Intelligence’s Function In Chatbots
That Use NLP And SVM Algorithms." 2023 3rd International Conference on Advance Computing and
Innovative Technologies in Engineering (ICACITE). IEEE, 2023.
[19] Wongpatikaseree, Konlakorn, et al. "Daily Health Monitoring Chatbot with Linear Regression." 2020
15th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-
NLP). IEEE, 2020.

© East Journal of Engineering 16

 Volume 1, Issue 1
East Journal of Engineering Publisher: East Publication & Technology
DOI: https://doi.org/
ISSN: 3079-9376

[20] Thamilselvan, R., et al. "Developing an AI-Driven Chatbot for Enhanced College Website Support
Using Machine Learning." 2024 International Conference on Expert Clouds and Applications (ICOECA).
IEEE, 2024.
[21] Nordheim, Cecilie Bertinussen, Asbjørn Følstad, and Cato Alexander Bjørkli. "An initial model of
trust in chatbots for customer service—findings from a questionnaire study." Interacting with
Computers 31.3 (2019): 317-335.
[22] Regin, R., S. Suman Rajest, and T. Shynu. "An automated conversation system using natural language
processing (nlp) chatbot in python." Central Asian Journal of Medical and Natural Science 3.4 (2022): 314-
336.
[23] Lalwani, Tarun, et al. "Implementation of a Chatbot System using AI and NLP." International Journal
of Innovative Research in Computer Science & Technology (IJIRCST) Volume-6, Issue-3 (2018).
[24] Khan, Shahnawaz, and Mustafa Raza Rabbani. "Artificial intelligence and NLP-based chatbot for
islamic banking and finance." International Journal of Information Retrieval Research (IJIRR) 11.3
(2021): 65-77.
[25] Ayanouz, Soufyane, Boudhir Anouar Abdelhakim, and Mohammed Benhmed. "A smart chatbot
architecture based NLP and machine learning for health care assistance." Proceedings of the 3rd
international conference on networking, information systems & security. 2020.
[26] Abdulla, Hussam, et al. "Chatbots development using natural language processing: A review." 2022
26th International Conference on Circuits, Systems, Communications and Computers (CSCC). IEEE, 2022.
[27] Husak, Vitaliy, et al. "Information System for Recommendation List Formation of Clothes Style Image
Selection According to User's Needs Based on NLP and Chatbots." COLINS 2604 (2020): 788-818.
[28] Mawardi, Viny Christanti. "Modified K-Nearest Neighbor Web Based Chatbot as Service and
Academic Information School." International Journal of Application on Sciences, Technology and
Engineering 1.2 (2023): 755-769.
[29] Bhasker, Mr PVM Vijay, et al. "Chat Bot Design Using NLP & KNN Algorithm."
[30] Nuruzzaman, Mohammad, and Omar Khadeer Hussain. "A survey on chatbot implementation in
customer service industry through deep neural networks." 2018 IEEE 15th international conference on e-
business engineering (ICEBE). IEEE, 2018.
[31] Mutiwokuziva, Milla T., et al. "A neural-network based chat bot." 2017 2nd International Conference
on Communication and Electronics Systems (ICCES). IEEE, 2017.
[32] Vamsi, G. Krishna, Akhtar Rasool, and Gaurav Hajela. "Chatbot: A deep neural network based human
to machine conversation model." 2020 11th International Conference on Computing, Communication and
Networking Technologies (ICCCNT). IEEE, 2020.
[33] Bhartiya, Namrata, et al. "Artificial neural network based university chatbot system." 2019 IEEE
Bombay Section Signature Conference (IBSSC). IEEE, 2019.
[34] Yu, Zhun, et al. "A decision tree method for building energy demand modeling." Energy and
Buildings 42.10 (2010): 1637-1646.
[35] Quinlan, J. Ross. "Induction of decision trees." Machine learning 1 (1986): 81-106.
[36] Safavian, S. Rasoul, and David Landgrebe. "A survey of decision tree classifier methodology." IEEE
transactions on systems, man, and cybernetics 21.3 (1991): 660-674.
[37] Breiman, Leo. Classification and regression trees. Routledge, 2017.

© East Journal of Engineering 17

 Volume 1, Issue 1
East Journal of Engineering Publisher: East Publication & Technology
DOI: https://doi.org/
ISSN: 3079-9376

[38] Hastie, Trevor. "The elements of statistical learning: data mining, inference, and prediction." (2009).
[39] Loh, Wei‐Yin. "Classification and regression trees." Wiley interdisciplinary reviews: data mining and
knowledge discovery 1.1 (2011): 14-23.
[40] Xydas, E. S., et al. "Forecasting electric vehicle charging demand using support vector
machines." 2013 48th international universities' power engineering conference (UPEC). IEEE, 2013.
[41] Cortes, Corinna. "Support-Vector Networks." Machine Learning (1995).
[42] Schölkopf, B. "Learning with kernels: support vector machines, regularization, optimization, and
beyond." (2002).
[43] Bishop, Christopher M. "Pattern recognition and machine learning." Springer google schola 2 (2006):
1122-1128.
[44] Manning, Christopher, and Hinrich Schutze. Foundations of statistical natural language processing.
MIT press, 1999.
[45] Okonkwo, Chinedu Wilfred, Lateef Babatunde Amusa, and Hossana Twinomurinzi. "COVID-Bot, an
intelligent system for COVID-19 vaccination screening: design and development." JMIR formative
research 6.10 (2022): e39157.
[46] Jurafsky, Daniel. "Speech and language processing." (2000).
[47] Brill, Eric. "Transformation-based error-driven learning and natural language processing: A case study
in part-of-speech tagging." Computational linguistics 21.4 (1995): 543-565.
[48] Thomas, Anu, and S. Sangeetha. "Deep learning architectures for named entity recognition: A
survey." Advanced Computing and Intelligent Engineering: Proceedings of ICACIE 2018, Volume 1.
Springer Singapore, 2020.
[49] Mikolov, Tomas, et al. "Distributed representations of words and phrases and their compositionality."
Advances in sneural information processing systems 26 (2013).
[50] A. Almasri, J. Alsaraireh, D. Salman and I. Aburagaga, "Explainable Artificial Intelligence Models
using Students' Academic Record Data, Tree Family Classifiers, and K-means Clustering to Predict
Students' Performance," 2022 10th International Conference on Smart Grid (icSmartGrid), Istanbul,
Turkey, 2022, pp. 46-51.

© East Journal of Engineering 18