JIGJIGA UNIVERSITY

INSTITUTE OF TECHNOLOGY
DEPARTMENT OF SOFTWARE ENGINEERING

UNDERGRADUATE PROJECT REPORT

Title: Diabetes Prediction System Using Machine Learning

I
Submitted By:

No. Name ID

1 Abdi Biranu R/0025/13

2 Merera garoma R/2045/13

3 Metasebia merkin R/2079/13

4 Idosa Tariku R/1618/13

5 Noah Tadase R/2388/13

6 Umalkheryat Saladin R/2884/13

7 Rukiya Mohammed R/2490/13

Advisor: Milion F
Signature: _____________________
Date: 12/Fab/2025
A PROJECT REPORT SUBMITTED TO THE DEPARTMENT OF
SOFTWARE ENGINEERING
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
AWARD OF THE DEGREE
OF BACHELOR OF SCIENCE IN SOFTWARE ENGINEERING
JIGJIGA UNIVERSITY
INSTITUTE OF TECHNOLOGY
DEPARTMENT OF SOFTWARE ENGINEERING
ACADEMIC YEAR: 2025
APPROVED BY:
Examiner 1: ________________________
Head of Department: ________________________
Date: 12/Feb/2025

II
ACKNOWLEDGEMENT
First and foremost, we extend our heartfelt gratitude to Almighty God, whose endless
blessings, guidance, and wisdom have been our source of strength throughout this
journey. Without His divine grace, the successful completion of this project would not
have been possible.
We are deeply indebted to our advisor, Milion F, for their invaluable guidance,
continuous support, and insightful feedback. Their expertise and encouragement have
played a crucial role in shaping this work, and we are truly grateful for their patience
and dedication.
Our sincere appreciation also goes to Jigjiga University, the Institute of Technology,
and the Department of Software Engineering for providing us with the necessary
resources, knowledge, and an enabling environment to undertake this project. The
faculty members, administrative staff, and fellow students have all contributed to our
academic and professional growth.
We would like to express our gratitude to our family and friends, whose unwavering
support, encouragement, and understanding have been a constant source of motivation.
Their belief in us has strengthened our resolve to persevere through challenges and
complete this project successfully.
Finally, we acknowledge all individuals, researchers, and developers whose work has
inspired and contributed to the development of this project. Their contributions in the
field of machine learning and healthcare technology have provided us with a solid
foundation to build upon.
To all who have played a part in this journey, we extend our heartfelt thanks. Your
support and encouragement have been instrumental in bringing this project to fruition.

III
Table of Contents
ACKNOWLEDGEMENT ........................................................................................... III

List of Figures ...................................................................................................... VIII

List of Tables .......................................................................................................... IX

List of Abbreviations .............................................................................................. IX

Chapter One: Introduction ............................................................................................. 1

1.1 Overview ................................................................................................................. 1

1.2 Statement of the Problem .................................................................................... 2

1.2.1 Existing System ...........................................................................................3

1.2.2 Major Problems of the Existing System ......................................................3

1.2.3 Proposed System ......................................................................................... 3

1.2.4 Advantages of the Proposed System ........................................................... 4

1.3 Motivation ...................................................................................................... 4

1.4 Scope and Limitations of the Project ............................................................. 5

1.4.1 Scope of the Project ................................................................................ 5

1.4.2 Limitations of the Project ....................................................................... 5

1.5 Objective ............................................................................................................. 6

1.5.1 General Objective ........................................................................................6

1.6.1 Data Collection Methodology .......................................................................... 9

1.6.1.1 Data Sourcing and Collection from Hospitals and Healthcare Facilities . 9

1.6.2 Data Cleaning and Preprocessing ...................................................................10

1.6.2.1 Handling Missing Data ...........................................................................10

1.6.2.2 Data Normalization and Standardization ............................................... 10

1.6.2.3 Outlier Detection and Removal ..............................................................10

1.6.2.4 Feature Selection and Engineering .........................................................10

IV
 1.6.2.5 Data Augmentation (if required) ............................................................ 10

1.6.2.6 Data Splitting ..........................................................................................11

1.6.3 System Design and Analysis Tools ................................................................11

1.6.3.1 System Architecture ............................................................................... 11

1.6.3.2 Use Case and System Flow Design ........................................................11

1.6.3.3 Tools for System Analysis and Design .................................................. 11

1.6.4 System Development Tools ........................................................................... 12

1.6.4.1 Front-End Development Tools ...............................................................12

1.6.4.2 Back-End Development Tools ............................................................... 12

1.6.4.3 Machine Learning Tools ........................................................................ 12

1.6.4.4 Database Tools ....................................................................................... 12

1.6.4.5 Deployment and Cloud Hosting ............................................................. 13

1.6.4.6 System Flowchart ................................................................................... 14

Figure 1.1 flowchart to visualize the data collection and model training
process. .......................................................................................................... 14

Chapter Two: System Requirement Specification .......................................................15

2.1 Background Overview ...................................................................................... 15

2.2 Functional Requirements .................................................................................. 17

2.2.1 User Registration and Authentication ....................................................... 18

2.2.2 Data Input and Collection ..........................................................................18

2.2.3 Predictive Modeling and Risk Assessment ............................................... 19

2.2.4 Display Results and Reporting .................................................................. 20

2.2.5 History Tracking and Data Storage ........................................................... 20

Key Functionalities: ........................................................................................... 21

2.2.6 System Maintenance and Updates .............................................................22

V
 Key Functionalities: ........................................................................................... 22

Types of Testing: ................................................................................................22

2.3 Non-Functional Requirements .......................................................................... 23

2.3.1 Performance Requirements ....................................................................... 23

Performance Requirements ................................................................................ 23

Key Considerations: .......................................................................................23

2.3.2 Scalability Requirements ...........................................................................24

2.3.3 Security Requirements .............................................................................. 24

2.3.4 Usability Requirements ............................................................................. 25

2.3.5 Reliability Requirements ...........................................................................26

2.3.6 Maintainability Requirements ................................................................... 26

2.3.7 Compatibility Requirements ......................................................................27

Key Considerations: .......................................................................................27

Benefits: .........................................................................................................28

2.4 Feasibility Study ...........................................................................................28

Long-Term Sustainability ..............................................................................29

2.4.1 Technical Feasibility ................................................................................. 29

2.4.2 Economic Feasibility .................................................................................31

2.4.3 Operational Feasibility .............................................................................. 32

2.4.4 Legal Feasibility ........................................................................................ 33

2.4.5 Social Feasibility ....................................................................................... 34

Chapter Three: System Analysis and Modeling .......................................................... 37

3.1 Overview ........................................................................................................... 37

3.2 Scenario-Based Modeling ................................................................................. 38

3.2.1 Use Case Identification ............................................................................. 38

VI
 Use Cases for the Diabetes Prediction System: ........................................ 38

Figure 3 Use Cases for the Diabetes Prediction System: ............................. 40

3.2.2 Actor Identification ................................................................................... 41

3.2.3 Use Case Description ................................................................................ 41

Use Case 1: Register Patient Information .................................................41

3.2.4 Activity Diagram .......................................................................................42

3.3Behavioral/Dynamic Modeling ..........................................................................44

3.3.1 Sequence Diagram .....................................................................................44

Figure 3.2 Sequence Diagram ...................................................................44

Figure 3.3: State Diagram ......................................................................... 46

Steps: .................................................................................................................. 47

Transitions: .........................................................................................................47

3.4 Class-Based Modeling ...................................................................................... 48

3.4.1 Identifying Classes .................................................................................... 48

3.4.2 Class Diagram ........................................................................................... 51

Explanation: ...................................................................................................52

Summary: .......................................................................................................53

Chapter 4: Conclusion and Future Work ..................................................................... 54

4.1 Conclusion .........................................................................................................54

4.2 Future Work ...................................................................................................... 54

Appendices ......................................................................................................... 56

Appendix A: Survey-Based Data ....................................................................... 56

Appendix C: Diagrams .................................................................................. 57

Appendix D: Dataset Description ..................................................................58

Appendix E: User Guide ................................................................................58

VII
 Appendix F: Ethical Considerations ..............................................................59

References .......................................................................................................... 60

List of Figures

Figure
Title/Description Section/Page
No.

Figure 1.1 Data Collection and Model Training Process Flowchart 14

Figure 3. Usecase Diagram for Data Submission 40

Figure 3.1 Activity Diagram for Data Submission 43

Figure 3.2 Sequence Diagram for Predict Diabetes Risk 44

Figure 3.3 State Diagram for Patient Data Processing 46

Figure 3.5 Class-Based Modeling - Patient 48

Figure 3.6 Class-Based Modeling - Prediction 49

Figure 3.7 Class-Based Modeling - Database 49

Figure 3.8 Class-Based Modeling - User Interface 49

Figure 3.9 Class Diagram Representation 52

VIII
List of Tables

Table No. Title/Description Section/Page

Table 1.1 Device Compatibility for Web Platform Chapter 1.5.2 (Specific Objectives)

Table 2.1 Infrastructure Requirements Chapter 2.4.1 (Technical Feasibility)

Table 2.2 Risk Factors & Mitigation Chapter 2.4.1 (Technical Feasibility)

Table 2.3 Financial Sustainability Strategies Chapter 2.4.2 (Economic Feasibility)

Table 2.4 Data Privacy & Security Chapter 2.4.3 (Operational Feasibility)

Table 2.5 Regulatory Compliance Chapter 2.4.4 (Legal Feasibility)

Table B1 Test Cases for Diabetes Prediction System Appendix B

List of Abbreviations

AI Artificial Intelligence

BMI Body Mass Index

CNN Convolutional Neural Network

DFD Data Flow Diagram

EHR Electronic Health Record

GDPR General Data Protection Regulation

IX
AI Artificial Intelligence

HbA1c Hemoglobin A1c

HIPAA Health Insurance Portability and Accountability Act

IDDM Insulin-Dependent Diabetes Mellitus (Type 1)

KNN K-Nearest Neighbors

ML Machine Learning

NIDDM Non-Insulin Dependent Diabetes Mellitus (Type 2)

RNN Recurrent Neural Network

SVM Support Vector Machine

WHO World Health Organization

X
Chapter One: Introduction

1.1 Overview

Diabetes is a group of metabolic disorders characterized by prolonged high blood

sugar levels due to abnormal insulin secretion or action. Common symptoms include
excessive urination, persistent thirst, and increased hunger. If left untreated, diabetes
can lead to severe complications such as diabetic ketoacidosis, hyperosmolar
hyperglycemic state, cardiovascular disease, stroke, kidney failure, foot ulcers, and
eye damage. It occurs when the pancreas fails to produce sufficient insulin or when
the body's cells do not respond properly to insulin. There are three main types of
diabetes. Type 1 Diabetes Mellitus, also known as Insulin-Dependent Diabetes
Mellitus, results from little to no insulin production by the pancreas and requires
external insulin administration for management. Type 2 Diabetes Mellitus, or Non-
Insulin Dependent Diabetes Mellitus, is characterized by insulin resistance, where the
body's cells do not respond effectively to insulin, and it is strongly associated with
obesity and a sedentary lifestyle. Gestational diabetes develops during pregnancy and
increases the risk of complications for both the mother and baby. For a healthy
individual, blood glucose levels typically range between 70 to 99 mg/dL, while a
fasting glucose level exceeding 126 mg/dL indicates diabetes, and levels between 100
to 125 mg/dL suggest pre-diabetes, a high-risk state for developing Type 2 diabetes.
Several factors increase the likelihood of developing diabetes, including a Body Mass
Index above 25, a family history of diabetes, low HDL cholesterol levels below 40
mg/dL, prolonged high blood pressure, a history of gestational diabetes, polycystic
ovary syndrome, belonging to high-risk ethnic groups such as African American,
Native American, Hispanic American, and Asian-Pacific populations, being over the
age of 45, and leading a sedentary lifestyle. Recent studies, such as those conducted
by Smith et al. in 2020, have demonstrated the effectiveness of machine learning
techniques like logistic regression in diabetes prediction, and the PIMA dataset
remains a benchmark for training predictive models.

1
However, many existing systems lack integration with real-time web platforms,
limiting accessibility and usability. This project aims to bridge that gap by developing
a web-based machine learning model for diabetes prediction, facilitating early
diagnosis and intervention.

1.2 Statement of the Problem

Diabetes is a chronic disease affecting millions worldwide, leading to severe

complications such as heart disease, kidney failure, nerve damage, and vision loss.
Early detection and management are crucial for preventing these complications and
improving patients' quality of life. However, traditional diagnostic methods rely on
physical tests and laboratory analysis, which can be time-consuming, expensive, and
inaccessible, particularly in low-resource settings.

With advancements in machine learning (ML) and web-based technologies, there is

an opportunity to develop an automated and efficient diabetes prediction system.
Many existing models lack real-time accessibility, user-friendliness, and accuracy in
diagnosing diabetes risk based on multiple factors such as age, BMI, blood pressure,
glucose levels, and family history.

This study aims to develop a machine learning-based web application that predicts
diabetes risk with high accuracy using historical medical data. The system will
provide an accessible and efficient solution for individuals and healthcare
professionals to assess diabetes risk early and take preventive measures. By
leveraging machine learning algorithms, the proposed system can improve diagnostic
accuracy and assist in timely medical interventions, reducing the long-term impact of
diabetes on individuals and healthcare systems.

2
1.2.1 Existing System

Currently, diabetes diagnosis is performed through various methods. Laboratory tests

such as fasting blood sugar (FBS) and HbA1c tests are commonly used to detect
diabetes by measuring blood glucose levels. In addition to laboratory tests, manual
analysis plays a crucial role, where doctors assess patient history, symptoms, and test
results to determine the risk of diabetes. Some hospitals also utilize risk assessment
tools, which are simple calculators that estimate diabetes risk based on factors such as
age, Body Mass Index (BMI), and lifestyle habits. These traditional methods, while
effective, can be time-consuming and may not always provide early detection,
highlighting the need for more advanced predictive systems.

1.2.2 Major Problems of the Existing System

Despite their effectiveness, traditional methods of diabetes diagnosis present several

challenges. One major drawback is that they are time-consuming, as laboratory tests
require multiple visits and waiting periods for results. Additionally, frequent medical
tests can be expensive, making them less accessible, especially in low-income regions
where healthcare costs are a significant burden. Limited accessibility is another
concern, as many remote areas lack proper healthcare facilities and diagnostic
equipment, preventing timely diagnosis and intervention. Manual errors also pose a
risk since the accuracy of risk assessment tools relies on human evaluation, which can
be subjective and prone to inconsistencies. Furthermore, traditional methods lack
predictive capability, as they do not leverage historical data to forecast future diabetes
risks. These limitations highlight the need for more efficient, affordable, and
technology-driven approaches to diabetes prediction and diagnosis.

1.2.3 Proposed System

To address these challenges, this project proposes a web-based machine learning

system for diabetes prediction. The system will leverage machine learning algorithms
to analyze patient data and predict diabetes risk with high accuracy.

3
By providing an accessible web-based platform, users will be able to check their
diabetes risk from anywhere without the need for extensive medical visits. The system
will incorporate real-time analysis, processing user inputs and generating instant
results to facilitate early diagnosis. Additionally, it will offer data-driven insights,
helping individuals take preventive measures based on their risk levels.

1.2.4 Advantages of the Proposed System

The proposed system offers several advantages over traditional diagnostic methods. It
will enables faster diagnosis by providing instant risk assessments without requiring
laboratory test results. The approach is cost-effective as it reduces the need for
frequent and expensive medical tests. Increased accessibility is another key benefit,
as the web-based platform allows remote users to assess their diabetes risk
conveniently. Machine learning models improve accuracy by analyzing multiple risk
factors simultaneously, leading to more reliable predictions. Furthermore, the system
is automated and scalable, making it suitable for both individuals and healthcare
professionals for large-scale screening. By integrating modern technology with
diabetes prediction, this system aims to bridge the gap between traditional diagnosis
methods and technological advancements, making early detection more efficient and
widely accessible.

1.3 Motivation

Diabetes, particularly Type 2 Diabetes Mellitus (T2DM), is a major global health

concern, affecting nearly 90% of diabetes patients. Research indicates that long non-
coding RNAs (lncRNAs) play a significant role in enhancing T2DM diagnosis,
highlighting the potential for advanced diagnostic methods. Machine learning
techniques provide powerful tools for analyzing, interpreting, and extracting
meaningful insights from medical data. These techniques can enhance early diagnosis
and prognosis, ultimately reducing the impact of diseases like T2DM.

4
1.4 Scope and Limitations of the Project

1.4.1 Scope of the Project

This project focuses on developing a web-based machine learning system for diabetes
prediction, with an emphasis on Type 2 Diabetes Mellitus (T2DM). The system will
utilize K-Nearest Neighbor (KNN), to enhance prediction accuracy. A user-friendly
web application will be developed, allowing users to input their health parameters and
receive an instant risk assessment. Machine learning algorithms will analyze key risk
factors such as age, BMI, glucose levels, blood pressure, and family history to provide
accurate predictions. The system will support data-driven decision-making, enabling
individuals to take preventive measures based on their assessed diabetes risk.
Additionally, the project is designed to be scalable, allowing for future integration of
additional health parameters and more advanced prediction techniques.

1.4.2 Limitations of the Project

Despite its advantages, the project has several limitations. The system is limited to
Type 2 Diabetes prediction and does not diagnose Type 1 Diabetes or Gestational
Diabetes. Its accuracy depends on the quality and completeness of input data,
meaning inaccurate or incomplete data may affect the results. The system does not
support real-time blood sugar monitoring, as it does not connect to medical devices
for continuous glucose level tracking. Additionally, the predictions are based on
machine learning models and should not be considered a replacement for professional
medical advice. The system requires an internet connection, making it inaccessible in
areas with poor connectivity. Privacy and security concerns must also be addressed to
ensure that patient data is securely stored and protected from unauthorized access.

1.6 Methodology

The methodology of this project follows a structured and systematic approach to

developing a machine learning-based diabetes prediction system. It encompasses data

5
collection, system design, and the development tools used to ensure the creation of an
accurate, secure, and efficient prediction model. This section also outlines the process
of gathering data from healthcare facilities to train the machine learning models for
diabetes prediction.

1.5 Objective

1.5.1 General Objective

The general objective of this project is to develop a machine learning-based web

application for predicting Type 2 Diabetes Mellitus (T2DM) with high accuracy.

[1.5.2 Specific Objectives]

To achieve the general objective, the project focuses on the following specific
objectives:

Develop a web-based platform for diabetes prediction:

Design and implement an intuitive and interactive web application where users can
input health data and receive instant feedback on diabetes risk.
Develop a web-based platform for diabetes prediction by following a systematic
lifecycle from data collection to deployment and testing:
Data Collection and Preparation:
Source and curate validated medical datasets (e.g., PIMA Indian Diabetes Dataset,
NHANES) containing features like age, BMI, glucose levels, and family history.
Preprocess data by addressing missing values (imputation), normalizing features (e.g.,
Min-Max scaling), and balancing class distributions (e.g., SMOTE for oversampling).
Model Development and Validation:
Train machine learning models (KNN, SVM, Random Forest, Logistic Regression)
using scikit-learn or TensorFlow, prioritizing interpretability and performance.
Optimize models through feature selection (e.g., Recursive Feature Elimination, PCA)
and hyperparameter tuning (e.g., GridSearchCV).
Validate models using stratified k-fold cross-validation and benchmark performance
against metrics like accuracy (>90%), recall (>85%), and AUC-ROC.

6
Web Application Design and Development:
Build an intuitive frontend using React.js for dynamic user interactions (e.g., input
forms, risk feedback).
Develop a backend API with Express or Node Js to process user inputs, run model
predictions, and return results in real time.
Integrate explainable AI (XAI) components (e.g., SHAP values, LIME) to visualize
risk factor contributions.
Cross-Device Compatibility and Responsiveness:
Implement responsive design frameworks (e.g., Bootstrap, CSS Grid) to ensure
seamless functionality across devices (see Table 1.1).
Conduct compatibility testing on desktops, tablets, and mobile devices to verify UI
adaptability and touch-friendly interfaces.
Deployment and Scalability:
Deploy the platform on cloud services (e.g., AWS, Heroku) with containerization
(Docker) for scalability.
Set up continuous integration/continuous deployment (CI/CD) pipelines using GitHub
Actions for automated updates.
System Testing and Usability Evaluation:
Perform functional testing (unit, integration, end-to-end) to validate prediction
accuracy and system robustness.
Conduct usability testing with stakeholders (patients, doctors) to refine UI/UX based
on feedback (e.g., simplifying input forms).
Security and Compliance:
Encrypt user data at rest (AES-256) and in transit (TLS 1.3).
Ensure compliance with healthcare regulations (HIPAA, GDPR) and implement
OAuth 2.0 for secure authentication.

Maintenance and Updates:

Monitor system performance post-deployment and retrain models periodically with
new data.
Provide user support channels (e.g., FAQs, chatbots) and update the platform to

7
address emerging user needs.
Ensure cross-device compatibility and responsiveness
Table 1.1: Device Compatibility

Device Type Compatibility Remarks

Desktop Fully Supported Optimized for large screens and browsers

Tablet Fully Supported Responsive design for varying screen sizes

Mobile Phone Fully Supported Mobile-friendly UI with touch interactions

Implement machine learning algorithms for accurate prediction:

Train and evaluate models (KNN, SVM, Random Forest, Logistic Regression) using
scikit-learn or TensorFlow, prioritizing interpretability and performance.
Optimize models via feature selection techniques (e.g., Recursive Feature Elimination,
PCA) to balance accuracy and computational efficiency.
Utilize medical datasets for training and validation:
Source real-world datasets (e.g., PIMA Indian Diabetes Dataset, NHANES) and
preprocess them by addressing missing values, scaling features, and balancing classes.
Split data into training (70%), validation (15%), and testing (15%) sets to ensure
robust model evaluation.
Analyze key risk factors associated with Type 2 Diabetes:
Incorporate factors like age, BMI, glucose levels, and family history, validating their
relevance through statistical analysis (e.g., logistic regression, SHAP values).
Generate visualizations (e.g., heatmaps, bar charts) to highlight correlations between
risk factors and diabetes likelihood.

Enhance user experience and system usability:

Conduct usability testing with stakeholders (patients, doctors) to refine the UI/UX.

8
Provide interactive dashboards with explainable AI (XAI) outputs to clarify
predictions.
Ensure security and privacy of user data:
Encrypt data at rest (AES-256) and in transit (TLS 1.3), and comply with
HIPAA/GDPR regulations.
Implement OAuth 2.0 for secure authentication and role-based access control.
Validate and compare system performance:
Benchmark models using metrics (accuracy, F1-score, AUC-ROC) and compare
against existing tools (e.g., ADA Risk Test).
Target >90% accuracy and >85% recall to minimize false negatives.
Provide actionable recommendations based on predictions:
Deliver personalized advice (e.g., diet plans, exercise routines) via the platform.
Enable API integration with healthcare providers for seamless referral systems.

1.6.1 Data Collection Methodology

Accurate diabetes predictions rely heavily on high-quality data. The data collection
process follows a systematic approach to ensure reliability and compliance with
ethical and legal standards.

1.6.1.1 Data Sourcing and Collection from Hospitals and Healthcare Facilities

The primary dataset for this project is obtained from hospitals, clinics, and healthcare
institutions to enhance real-world applicability. The process involves:

Formal Agreements and Data Access

 Obtain formal agreements with healthcare institutions for access to patient

data.

 Ensure compliance with data privacy regulations such as HIPAA (Health

Insurance Portability and Accountability Act) and GDPR.

Health Records

9
  Collect electronic health records (EHRs) containing key diabetes risk factors:

 Age, gender, weight, BMI, glucose levels, blood pressure, family history, and
physical activity.

 Anonymize data by removing personally identifiable information (PII) to

protect patient privacy.

1.6.2 Data Cleaning and Preprocessing

After data collection, preprocessing ensures the dataset is suitable for machine
learning training.

1.6.2.1 Handling Missing Data

 Use techniques such as mean/mode imputation, interpolation, or removal of

incomplete records.

1.6.2.2 Data Normalization and Standardization

 Normalize or standardize numerical features like blood glucose levels, BMI,

and blood pressure for consistency.

1.6.2.3 Outlier Detection and Removal

 Identify and remove outliers that could negatively impact model performance.

1.6.2.4 Feature Selection and Engineering

 Select relevant features using correlation analysis and domain knowledge.

 Apply feature engineering techniques such as BMI categorization and glucose
level classification to improve model accuracy.

1.6.2.5 Data Augmentation (if required)

 If the dataset is imbalanced, use SMOTE (Synthetic Minority Over-

10
 sampling Technique) or other data augmentation methods to balance it.

1.6.2.6 Data Splitting

 Training Set (80%) – Used for training machine learning models.

 Testing Set (20%) – Used to evaluate the model’s accuracy and
generalization.

1.6.3 System Design and Analysis Tools

The system is designed using a combination of tools and frameworks to ensure an

intuitive user interface and efficient machine learning model integration.

1.6.3.1 System Architecture

 The system follows a client-server architecture, where users interact through

a web interface to input health data.
 The backend processes the data using machine learning models to generate
predictions.

1.6.3.2 Use Case and System Flow Design

 Use Case Diagrams define system functionalities, including user registration,

health data input, prediction model execution, and result visualization.
 System Flow Diagrams illustrate the process from data input to machine
learning model prediction and result output.

1.6.3.3 Tools for System Analysis and Design

 UML (Unified Modeling Language): Used for designing system components

and relationships between different entities (users, machine learning models,
etc.).
 Data Flow Diagrams (DFD): Used to visualize how data moves through the
system from input collection to prediction output.

11
1.6.4 System Development Tools

The system is developed using industry-standard programming languages,

frameworks, and databases.

1.6.4.1 Front-End Development Tools

 React.js – Builds an interactive, responsive web interface for user input,

predictions, and result display.
 HTML, CSS, JavaScript – Ensure a visually appealing and functional user
interface.

1.6.4.2 Back-End Development Tools

 Node.js – Handles API requests, data processing, and machine learning model
integration.
 Express.js – A web framework for structuring the backend server.
 Python – Used for machine learning model training, data processing, and
evaluation.
 scikit-learn: Implements classification models (KNN, SVM, Random Forest,
Logistic Regression).

1.6.4.3 Machine Learning Tools

 Python (NumPy, Pandas, Scikit-learn) – Used for data manipulation, model

training, and evaluation.
 Jupyter Notebook – Enables experimenting with different machine learning
models and visualizing data.
 TensorFlow/Keras (if needed) – For advanced deep learning models and
optimizations.

1.6.4.4 Database Tools

 PostgreSQL/MySQL – Stores patient health data securely in a relational

database.
 MongoDB (if necessary) – A NoSQL database for handling large, flexible

12
 datasets.

1.6.4.5 Deployment and Cloud Hosting

 AWS, Google Cloud, or Heroku – Cloud platforms for deploying the web
application.
 Git/GitHub – Version control and collaborative development.

13
1.6.4.6 System Flowchart

Figure 1.1 flowchart to visualize the data collection and model training
process.

14
Chapter Two: System Requirement Specification
2.1 Background Overview

Diabetes is one of the leading global public health challenges of the 21st century. This
chronic metabolic disorder is characterized by elevated levels of blood glucose due to
issues related to insulin production or its effectiveness. According to the World
Health Organization (WHO), an estimated 463 million people were living with
diabetes in 2019, and this number is expected to rise significantly in the coming
decades. The disease is not only a personal health issue but also a public health
burden, straining healthcare systems worldwide and contributing to a variety of long-
term complications, such as heart disease, stroke, kidney failure, blindness, and
amputation.

Type 2 Diabetes Mellitus (T2DM), which accounts for approximately 90% of all
diabetes cases, is a chronic condition where the body either becomes resistant to
insulin or doesn’t produce enough insulin to maintain normal glucose levels.
Although it is often preventable, many individuals develop T2DM without early
symptoms. This makes early detection and intervention vital to prevent the onset of
more severe health complications. Early-stage diabetes or pre-diabetes can often be
managed through lifestyle changes such as improved diet and increased physical
activity, making the role of early detection even more critical.

Despite the availability of traditional diagnostic methods such as blood glucose tests
and oral glucose tolerance tests, late diagnosis continues to be a significant challenge.
Many individuals with diabetes remain undiagnosed, and by the time they are
diagnosed, they often have already developed severe complications. The delay in
diagnosis is often linked to the absence of early symptoms in the case of Type 2
diabetes and the lack of regular screening for individuals at risk.

In response to this challenge, the integration of machine learning (ML) and predictive
analytics in healthcare offers an innovative solution. Machine learning algorithms,

15
with their ability to analyze large datasets and recognize patterns in complex data,
have proven effective in identifying early warning signs of various diseases, including
diabetes. These algorithms can process diverse health data, including age, gender,
BMI (Body Mass Index), family history, physical activity, and diet, to predict the
likelihood of diabetes onset before the disease progresses to more serious stages.

The healthcare sector is witnessing an increasing reliance on electronic health records

(EHRs), which house a wealth of patient data. This data, combined with the
development of machine learning models, forms a strong foundation for building a
predictive diabetes risk assessment system. With access to vast amounts of clinical
data, machine learning models can be trained to classify individuals into risk
categories, enabling healthcare providers to identify those who are likely to develop
diabetes in the future. This can lead to earlier preventive interventions, personalized
care plans, and better management of pre-diabetic conditions.

One of the major advantages of using machine learning for diabetes prediction is that
it can provide real-time analysis of patient data, making it a useful tool for healthcare
providers. In many cases, a web-based platform can be developed to integrate with
existing healthcare infrastructure, offering doctors and healthcare professionals a tool
that provides fast and accurate predictions. This system would use historical patient
data, processed through machine learning algorithms, to make predictions about an
individual’s risk of developing diabetes.

Additionally, the increasing adoption of wearable health technology and mobile

health applications has created new avenues for gathering data, such as daily physical
activity and blood glucose levels, which can be incorporated into the prediction
models. This provides the opportunity for continuous monitoring and predictive
analysis even outside the clinical setting, offering a more holistic approach to diabetes
management.

16
Moreover, predictive diabetes systems can be tailored to meet the needs of specific
populations, such as high-risk individuals with a family history of diabetes, elderly
patients, or individuals with obesity. This customization of predictions can make the
system more relevant and accurate, ensuring that at-risk individuals are identified
early and provided with targeted interventions.

In the current technological landscape, web-based diabetes prediction systems offer

the potential to transform the way diabetes risk is assessed, diagnosed, and managed.
By leveraging machine learning techniques, these systems can provide healthcare
professionals with an easy-to-use, cost-effective, and scalable solution to improve
both the efficiency and effectiveness of diabetes care.

As such, the goal of this project is to create a web-based diabetes prediction system
that utilizes machine learning models, specifically classification algorithms, to predict
the likelihood of an individual developing diabetes based on a variety of health-
related factors. The system will be designed with a user-friendly interface that can be
easily integrated into existing healthcare workflows, allowing healthcare professionals
to use it as a tool for early diagnosis and preventive healthcare.

In addition to its diagnostic potential, this system aims to contribute to the growing
field of digital health technologies, which are revolutionizing the way healthcare is
delivered. By improving predictive accuracy, reducing diagnostic errors, and offering
real-time analysis, the proposed system has the potential to reduce the burden of
diabetes on individuals and healthcare systems globally

2.2 Functional Requirements

Functional Requirements describe the specific behavior or functions of the system

that must be implemented to satisfy the needs of the users and stakeholders. In this
section, we outline the key functionalities of the Diabetes Prediction System. The
system should be able to predict the likelihood of diabetes based on user input and
provide healthcare professionals with relevant information to support decision-making.

17
The functional requirements cover all the core features and operations that the system
must support for it to be functional and useful to the healthcare professionals using it.

2.2.1 User Registration and Authentication

The Diabetes Prediction System must support a secure and efficient user
registration and authentication process for authorized individuals, such as
healthcare professionals. Users should be able to create new accounts by
providing personal details like username, email, password, and role (e.g.,
doctor, nurse). The system will validate the user’s credentials during login and
provide feedback in case of incorrect information. Additionally, password
recovery functionality will be available for users who forget their credentials.
User access will be role-based to ensure that only authorized personnel can
access sensitive data and predictive tools.
User Access Levels:

 Patients: Patients can input their personal health data and view general
recommendations, but they do not have access to raw prediction results or
sensitive healthcare information.
 Healthcare Providers: Healthcare professionals, such as doctors and nurses,
can enter and review patient data, interpret predictions, and generate reports to
support decision-making.
 Administrators: Administrators are responsible for managing user accounts,
monitoring system performance, and ensuring the overall security of the
system, including access control and data protection measures.

2.2.2 Data Input and Collection

The Diabetes Prediction System should enable healthcare professionals or patients to

input relevant personal and health-related data essential for diabetes risk prediction.
The system should have input fields for collecting various types of information,
including demographic details such as age, gender, family history of diabetes, and

18
ethnicity. It should also gather important health information, such as Body Mass
Index (BMI), blood pressure, cholesterol levels, glucose levels, physical activity
levels, and lifestyle factors like diet and smoking habits. Additionally, the system
should account for any existing symptoms related to diabetes, such as excessive thirst,
frequent urination, or blurred vision.
Data preprocessing should be carried out on the input data, such as normalizing
numerical values like BMI and glucose levels, ensuring they are in a consistent format
for accurate predictions. Input validation is crucial to maintain the accuracy of the
data, and the system should ensure that all inputs, such as BMI, blood pressure, and
glucose levels, are correctly formatted (e.g., numeric values) and handle missing data
appropriately by either providing default values or notifying the user to input the
missing information. This ensures reliable input and enhances the system's predictive
capability.

2.2.3 Predictive Modeling and Risk Assessment

The core functionality of the Diabetes Prediction System is to analyze the input data
and predict the likelihood of diabetes development using machine learning models. To
achieve this, the system should implement one or more classification algorithms, such
as Logistic Regression, Decision Trees, Random Forest, or K-Nearest Neighbors
(KNN), that are capable of analyzing the various risk factors input by users.
Based on the collected data, the system should calculate and provide a diabetes risk
score or a probability (e.g., 80% chance of developing Type 2 Diabetes). This
prediction is crucial for determining the user's level of risk for diabetes. Additionally,
the system should incorporate a pre-defined risk threshold, such as 70%, which
categorizes the user into one of three risk levels: high risk, medium risk, or low risk.

For individuals identified as high-risk, the system should offer decision support by
suggesting preventive actions, such as recommending lifestyle changes, more frequent
testing, or referring the patient to a specialist for further evaluation. For those
identified as low-risk, the system can recommend ongoing monitoring or regular
check-ups to ensure early detection and continued health management. This

19
functionality not only helps users assess their risk but also provides actionable
insights for improving or maintaining their health.

2.2.4 Display Results and Reporting

Once the prediction is made, the Diabetes Prediction System should display the
results in a clear and understandable format, enabling healthcare professionals to
interpret the findings effectively. The system will feature a Results Dashboard that
presents the diabetes risk assessment report. This report should include:
 The risk score or probability of developing diabetes, providing a clear
numerical representation of the individual's risk.
 A summary of input data, such as BMI, age, cholesterol levels, and other
health factors, displayed with visual indicators like charts or graphs for
easier interpretation.
 Recommendations based on the results, including suggestions for dietary
adjustments, exercise plans, or referrals to specialists, helping healthcare
professionals make informed decisions about the patient's care.

The User-Friendly Interface ensures that the results are easily understood by
medical professionals, using visual aids like bar charts, pie charts, and risk
category indicators (e.g., high, medium, low) to highlight important information.

Additionally, the system will offer a Downloadable Report feature, allowing

healthcare professionals to download the full report in formats like PDF or CSV. This
provides convenience for further analysis, record-keeping, or sharing the information
with patients or other medical staff.

2.2.5 History Tracking and Data Storage

The Diabetes Prediction System should have the capability to store and track
historical data for each user, enabling healthcare professionals to review past risk
assessments and monitor the patient's progress over time. This functionality ensures

20
that medical professionals can make data-driven decisions based on the patient’s
historical health trends.

Key Functionalities:

 Patient History: The system should securely store each patient’s data in a
database, which includes past risk predictions, health data (e.g., glucose levels,
BMI), and any treatment or preventive recommendations that were made. This
historical data will provide a comprehensive view of the patient's journey and
help in assessing changes in their health over time.

 Data Retrieval: Healthcare professionals should have the ability to easily

retrieve a patient's past reports, comparing them with current risk assessments.
This will help track improvements or deteriorations in the patient's health,
providing insights into the effectiveness of treatments or lifestyle changes.

 Update Data: The system should allow users (with appropriate permissions)
to update their personal health data, such as weight, blood pressure, or
cholesterol levels. These updates will be factored into future risk predictions,
ensuring that the system remains accurate and up-to-date based on the
patient’s current health status.

 Data Security: All stored data must be encrypted and protected from
unauthorized access, ensuring that sensitive health information is kept secure.
The system should adhere to health data security standards, such as HIPAA
(Health Insurance Portability and Accountability Act) in the U.S. or applicable
regulations in other regions, to ensure compliance with privacy laws.

21
2.2.6 System Maintenance and Updates

The system must be designed to support ongoing maintenance and periodic updates to
ensure it remains functional, secure, and capable of adapting to evolving healthcare
needs. This will allow the system to continue providing accurate and reliable
predictions over time while maintaining its performance and compliance with industry
standards.

Key Functionalities:

 Model Updates: The system should be able to integrate new machine

learning models and data sources periodically to improve the prediction
accuracy. This flexibility will allow the system to adapt to changes in
healthcare trends, data, and research, ensuring that predictions are based on the
most current information available.

 Bug Fixes: A continuous monitoring framework should be established to

identify and resolve any bugs or issues that arise during system use. This will
ensure the system remains reliable and functional, minimizing downtime and
maintaining a positive user experience.

 Security Patches: The system should be regularly updated with security

patches to address potential vulnerabilities. These updates are critical for
maintaining compliance with regulatory standards, such as HIPAA or GDPR,
and ensuring patient data is always protected against unauthorized access.

Types of Testing:

 Unit Testing: Tests individual components of the system, such as input

validation functions, to ensure they work correctly in isolation.
 Integration Testing: Ensures that the system's APIs and machine learning
models work seamlessly together and that data flows correctly between
components.

22
  User Acceptance Testing (UAT): Involves gathering feedback from
healthcare providers to validate the system’s usability and ensure that it
meets the needs of medical professionals. For example, "User Acceptance
Testing (UAT) will involve 10 healthcare professionals from Jigjiga General
Hospital to validate system usability."

2.3 Non-Functional Requirements

Non-functional requirements define the quality attributes, system performance, and

operational constraints that the system must meet. Unlike functional requirements,
which describe what the system will do, non-functional requirements describe how
the system will perform. These requirements are essential for ensuring that the system
is efficient, secure, scalable, and reliable. Below is a detailed explanation of the non-
functional requirements for the Diabetes Prediction System.

2.3.1 Performance Requirements

Performance Requirements

Performance requirements are essential to ensure that the Diabetes Prediction System
operates efficiently, especially when faced with high volumes of user interactions and
large datasets. The system should be designed to provide timely, accurate predictions
without compromising on responsiveness or usability.

Key Considerations:

 Response Time: The system must provide predictions within a maximum

time frame of 3 seconds for individual user inputs. This ensures that users
receive near-instantaneous feedback. For more complex queries or bulk data
processing, the system should still maintain efficiency, with no delays
exceeding 5 seconds. This will help keep users engaged and prevent

23
 frustration.

 Throughput: The system should be capable of handling may users

simultaneously without noticeable performance degradation. The backend
infrastructure must be scalable and robust enough to process prediction
requests concurrently while maintaining a smooth user experience. This
ensures that multiple healthcare professionals or patients can interact with the
system without delays or disruptions.

 Data Load: The system should be optimized to handle large datasets

efficiently. It must be capable of processing and storing health data for up to
Many users while maintaining high performance and low latency. The
database and backend infrastructure must be well-designed to accommodate
this volume, ensuring that the system remains responsive even as data grows.

2.3.2 Scalability Requirements

2.3.3 Security Requirements

Description: Security requirements are crucial to protect sensitive patient data and
ensure that the system is resilient to malicious attacks. Healthcare data is highly
sensitive, and thus, security must be a top priority in the system design.

Key Considerations:

Data Encryption: All sensitive data, including patient personal information and
medical data, must be encrypted both in transit and at rest. SSL/TLS encryption
should be used for all communication between the user and the server.

Authentication and Authorization: The system must implement robust user

authentication (e.g., using multi-factor authentication) and role-based access
control (RBAC) to ensure that only authorized users (e.g., healthcare professionals)
have access to patient data and system functionalities.

Data Integrity: The system must ensure the integrity of data using mechanisms like

24
checksums and hashing. This ensures that data cannot be altered or tampered with by
unauthorized parties.

Audit Logging: The system should keep an audit trail of all user actions, especially
for access to patient data, predictions, and system modifications. These logs should be
secure and accessible only to authorized personnel for compliance and monitoring.

Compliance with Regulations: The system must comply with relevant data
protection and healthcare regulations such as HIPAA (Health Insurance Portability
and Accountability Act) or GDPR (General Data Protection Regulation) to ensure
patient privacy and data protection.

2.3.4 Usability Requirements

Description: Usability requirements focus on how easy and intuitive the system is to
use for healthcare professionals and patients. The goal is to ensure that users can
interact with the system effectively without needing extensive training.

Key Considerations:

User Interface (UI): The user interface should be simple, intuitive, and responsive.
It must present relevant information clearly, using visual aids such as charts, graphs,
and risk indicators to aid in decision-making.
User Experience (UX): The system should follow best practices for UX design to
ensure that tasks such as data input, risk prediction, and report generation can be
performed with minimal effort and maximum efficiency.
Error Handling: The system should provide clear error messages and instructions
to help users resolve issues when input data is incorrect or missing.
Accessibility: The system should be designed to meet accessibility standards,
ensuring that it is usable by individuals with disabilities (e.g., support for screen
readers, color contrast for the visually impaired).

25
2.3.5 Reliability Requirements

Description: Reliability refers to the system’s ability to function consistently and

without failure over time. The system must be available for use when required and
capable of recovering from potential failures.

Key Considerations:

System Availability: The system should be available 99.9% of the time, ensuring
minimal downtime for both users and administrators. This should be backed by a
service level agreement (SLA).

Backup and Recovery: The system should include regular backups of patient data
and models. In case of failure, the system should be able to recover quickly with
minimal data loss (e.g., restore from the most recent backup within 15 minutes).

Fault Tolerance: The system should be designed to handle partial failures

gracefully. If one component (e.g., a server or database) fails, the system should still
be able to function with minimal impact on performance or user experience.

Error Detection and Reporting: The system should automatically detect errors and
log them for troubleshooting. Automatic alerts should notify system administrators if
critical failures occur.

2.3.6 Maintainability Requirements

Description: Maintainability refers to how easy it is to update, improve, and fix the
system over time. As technology and healthcare needs evolve, the system must be
adaptable and maintainable.

Key Considerations:

26
Modular Design: The system should follow a modular architecture to make it
easier to update individual components (e.g., machine learning models, user interface)
without affecting the entire system.

Code Quality: The system’s code should be well-documented, clean, and follow
coding best practices. This ensures that future developers can understand and maintain
the system efficiently.

Automated Testing: The system should include automated testing tools that allow
developers to quickly verify the functionality of the system after updates or bug fixes.

Version Control: The system should use version control systems (e.g., Git) to track
changes in code, ensuring that updates and modifications can be managed effectively.

2.3.7 Compatibility Requirements

Compatibility requirements ensure that the Diabetes Prediction System functions

effectively across various systems, platforms, and technologies used within healthcare
environments. These requirements are critical for ensuring seamless integration,
accessibility, and efficient operation across diverse devices and systems.

Key Considerations:

Cross-Browser Compatibility: The system's user interface must be compatible

with the most commonly used web browsers, including Google Chrome, Mozilla
Firefox, Safari, and Microsoft Edge. This ensures that healthcare professionals
and patients can access the system regardless of their preferred browser.

 Mobile Compatibility: The system should be responsive, meaning it can

adjust to different screen sizes and devices. It should work seamlessly on both
Android and iOS platforms, allowing healthcare professionals to access the
system remotely, whether using smartphones or tablets.

 Integration with External Systems: The Diabetes Prediction System must be

27
 able to integrate with other healthcare systems, particularly Electronic Health
Records (EHR), to ensure that patient data can be accessed and updated in
real-time. This integration will reduce the need for redundant data entry and
allow for a streamlined workflow across different healthcare platforms.

 Data Exchange Standards: To support smooth interoperability with other

healthcare systems, the Diabetes Prediction System should follow standard
protocols for data exchange, such as HL7 (Health Level 7) or

FHIR (Fast Healthcare Interoperability Resources). These protocols will allow

the system to securely exchange data with other systems while maintaining
compatibility with existing healthcare infrastructure.

Benefits:

 Increased Accessibility: Ensuring cross-browser and mobile compatibility

enhances the system's accessibility for healthcare professionals, allowing them
to use the system in various environments, including on the go.
 Seamless Integration: Integration with EHR systems will help streamline data
flow between systems, reduce errors, and improve patient care.
 Future-Proofing: Adhering to established data exchange standards ensures
that the system remains compatible with future healthcare technologies and
platforms.

2.4 Feasibility Study

The Diabetes Prediction System aims to enhance early detection of diabetes using
machine learning technologies. A thorough feasibility study is conducted to evaluate
the practicality and sustainability of the system in terms of technical, economic,
operational, legal, and social factors. This comprehensive analysis ensures that the
system is viable for development, deployment, and long-term use.

28
Long-Term Sustainability

To ensure long-term sustainability, the Diabetes Prediction System will:

 Establish partnerships with healthcare providers for continuous data

collection and model improvements.
 Incorporate a scalable architecture that allows the system to expand to
predict other chronic diseases, thus maximizing its healthcare impact.

2.4.1 Technical Feasibility

Technical feasibility assesses whether the available technology and resources can
support the development and deployment of the Diabetes Prediction System.

Technology Stack The system will utilize the following technologies:

 Machine Learning: Python, Scikit-learn, TensorFlow, and Pandas for data

preprocessing and model training.
 Frontend: React.js for an interactive and responsive user interface.
 Backend: Node.js with Express.js for API request handling.
 Database: MongoDB for efficient patient data storage.
 Hosting: Cloud platforms (AWS, Google Cloud) for scalability and reliability.

Data Collection

 The system will initially use the PIMA Indian Diabetes Dataset for training.
 Healthcare partnerships will help collect real-world data to enhance model
accuracy.

Model Training

 The system will implement Logistic Regression, K-Nearest Neighbors

(KNN), and Random Forest algorithms to predict diabetes risk.
 Continuous model retraining will improve prediction accuracy based on new
patient data.9

29
Infrastructure Requirements (Table 2.1):

Component Requirement

RAM Minimum 8GB (Recommended: 16GB)

Processor Intel i5+ or equivalent with GPU

Storage 200GB+ for datasets and models

Cloud Deployment AWS, GCP, or Azure for scaling

Risk Factors & Mitigation (Table 2.2):

Risk Impact Mitigation Strategy

Model accuracy decline High Regular retraining with updated data

30
 Risk Impact Mitigation Strategy

System overload Medium Autoscaling cloud infrastructure

Data security threats High Encryption & role-based access control

2.4.2 Economic Feasibility

Economic feasibility evaluates the system’s cost-effectiveness and financial

sustainability.

Initial Development Costs

 Software development costs (developer salaries, infrastructure setup).

 Cloud hosting (AWS, Google Cloud).
 Data acquisition from healthcare providers.
 Use of open-source technologies (Python, React, Node.js) reduces overall
costs.

Ongoing Operational Costs

 Cloud service expenses (storage, computation, scaling).

 Machine learning model maintenance (updates and retraining).
 Security and compliance monitoring to ensure data protection.

Cost-Benefit Analysis

 Early diabetes detection reduces long-term healthcare costs by preventing

complications.
 Subscription-based models for hospitals generate consistent revenue.
 Collaboration with pharmaceutical companies for predictive analytics

31
 increases financial sustainability.

Financial Sustainability Strategies (Table 2.3):

Revenue Model Potential Benefit

Hospital subscriptions Recurring revenue from healthcare providers

Data analytics licensing Revenue from research institutions

Collaboration with insurance companies Monetization through preventive insights

2.4.3 Operational Feasibility

Operational feasibility evaluates how practical the system will be for daily use in
healthcare settings.

User Acceptance & Training

 Healthcare providers will receive training to ensure effective system use.

 The user interface (UI) will be designed for non-technical users, promoting
ease of use.

System Integration

 The system will integrate with Electronic Health Records (EHR) systems
using FHIR (Fast Healthcare Interoperability Resources) for compatibility.
 Web-based access enables healthcare providers to access the system remotely.

Data Privacy & Security (Table 2.4):

Security Measure Implementation

32
 Security Measure Implementation

HIPAA/GDPR Compliance Ensuring encrypted patient data storage

Role-Based Access Restricting access to authorized users

Multi-Factor Authentication (MFA) Enhancing security with additional login verification

Impact on Healthcare Operations

 The system will reduce manual diagnosis time by providing automated

diabetes risk predictions.
 It aids decision-making but does not replace medical expertise.
 Early intervention through risk predictions will reduce long-term treatment
costs.

2.4.4 Legal Feasibility

Legal feasibility ensures that the system complies with healthcare and data protection
laws.

Regulatory Compliance (Table 2.5):

Regulation Requirement

HIPAA (USA) Protects patient data privacy

GDPR (EU) Ensures data security & consent-based usage

FDA Approval May be required if classified as a medical tool

33
Intellectual Property & Data Rights

 The ML model, software code, and algorithms will be protected under

copyright and patent laws.
 Data-sharing agreements will define data ownership between hospitals and
the system.

2.4.5 Social Feasibility

Social feasibility assesses how well the system will be accepted by patients,
healthcare professionals, and the public.

Patient Trust & Transparency

 Explainable AI (XAI): Patients will be able to understand the reasoning

behind the diabetes risk prediction.
 Clear Reporting: Healthcare providers will receive a detailed breakdown of
predictions.

Healthcare Provider Adoption

 Initial skepticism from healthcare providers will be addressed through pilot

testing and evidence-backed results.
 User feedback loops will be established to improve system accuracy and
build trust.

34
Public Health Impact

 Early diabetes detection can reduce the incidence of complications such as

heart disease and kidney failure.
 This leads to improvements in quality of life and reduced long-term
healthcare costs.

35
36
Chapter Three: System Analysis and Modeling
3.1 Overview

The System Analysis and Modeling chapter is a crucial part of the documentation that
bridges the gap between the conceptual design and practical implementation of the
Diabetes Prediction System. It begins with Problem Analysis, where the core issue of
delayed diabetes detection is identified, highlighting the impact of late diagnoses on
healthcare costs, patient health, and overall quality of life. The system’s goal is to
provide early prediction and diagnosis of diabetes, enabling timely intervention.
Following this, the Requirements Revisited section revisits the system's functional
and non-functional requirements. Functional requirements focus on the system’s core
tasks—collecting patient data, processing it, and generating predictions, while non-
functional requirements emphasize the system’s performance, scalability, security,
and user-friendliness. The System Design section outlines the system’s architecture,
which includes the frontend (user interface), backend (data processing and machine
learning model), and cloud storage. User interface design prioritizes ease of use for
both healthcare professionals and patients, with wireframes and mockups illustrating
the flow of data and results. The backend design details the processes involved in data
handling, prediction logic, and communication with the machine learning model.
Database design outlines the structure of patient data, predictions, and historical
reports. In the Modeling section, various models and diagrams, such as use case
diagrams, data flow diagrams, entity-relationship diagrams, and class diagrams, are
used to visually represent system functionality and data flow. Finally, the System
Architecture and Interaction section explains how the system components interact,
focusing on frontend-backend communication, machine learning integration, and
cloud infrastructure. This chapter serves as a comprehensive blueprint for developing
a system that provides accurate and timely diabetes predictions, ultimately enhancing
patient care and reducing healthcare costs.

37
3.2 Scenario-Based Modeling

Scenario-Based Modeling focuses on understanding how the system will function in

real-world situations by identifying and defining the interactions between users
(actors) and the system. This approach helps break down the system’s functionality
through use cases and provides insight into how the system should behave during
various processes or scenarios. In this section, we'll define use cases, identify the
actors, provide use case descriptions, and model activity diagrams.

3.2.1 Use Case Identification

A use case is a description of a system’s functionality from the perspective of its

users. It illustrates what the system does in response to user actions, without
specifying how the system performs the actions.

Use Cases for the Diabetes Prediction System:

Register Patient Information:

1. Description: The user (healthcare provider or patient) enters basic
information such as name, age, gender, family history, lifestyle factors,
and medical history.
2. Goal: To collect essential data for processing diabetes risk.

Submit Patient Data for Analysis:

1. Description: After registration, the user submits health-related data

(e.g., glucose levels, BMI, blood pressure) to the system for prediction.
2. Goal: To analyze the input data and generate a risk prediction.

38
Predict Diabetes Risk:

1. Description: The system processes the submitted data using the

machine learning model to predict whether the patient is at risk of
developing diabetes.
2. Goal: To evaluate the likelihood of diabetes and inform the user about
the prediction result.

View Diabetes Prediction Result:

1. Description: After prediction, the system provides the user with a

report detailing the risk of diabetes and recommendations for lifestyle
changes.
2. Goal: To present the prediction outcome and guide the user on the next
steps.

Update Patient Information:

1. Description: The healthcare provider or patient updates any new

health information, such as changes in medical condition or lifestyle
habits.
2. Goal: To ensure that the system’s prediction is based on up-to-date
information.

Generate Reports:

1. Description: The system generates detailed reports of diabetes

predictions, trends, and patient data.
2. Goal: To allow healthcare providers to monitor and review patient data
over time.

39
Figure 3 Use Cases for the Diabetes Prediction System:

40
3.2.2 Actor Identification

An actor in a system refers to any entity (person, external system, or device) that
interacts with the system. For the Diabetes Prediction System, the primary actors are:

Patient:

A user who provides personal health information, submits data for

analysis, and views the results.

Healthcare Provider:

A doctor or medical professional who registers patient data, reviews

predictions, and generates reports for further medical action.

Admin:

Responsible for system maintenance, user management, and ensuring

that data is correctly handled and stored.

Machine Learning Model:

This is a system actor that processes the patient’s data and provides
predictions based on the trained model.

System:

The system itself, which performs calculations, provides results,

generates reports, and manages interactions between the actors.

3.2.3 Use Case Description

For each use case identified, a detailed use case description can be provided,
including the steps involved, preconditions, and postconditions. Below is an example
of a Use Case Description:

Use Case 1: Register Patient Information

 Use Case Name: Register Patient Information

 Actor: Healthcare Provider

41
  Description: The healthcare provider enters the patient’s basic details into the
system for diabetes prediction.
 Preconditions: The healthcare provider has access to the system. Patient
details are available.
 Basic Flow:

1. The healthcare provider logs into the system.

2. The system displays the "Register Patient" form.
3. The healthcare provider enters patient information such as name, age,
medical history, and family history.
4. The healthcare provider submits the form.
5. The system stores the patient’s details in the database.

 Postconditions: The patient’s information is stored in the system and is

available for data submission.
 Alternative Flows:

o If the healthcare provider enters incomplete information, the system

prompts them to fill in the missing details.

 Exceptions: If the patient’s details cannot be saved due to a system failure, the
healthcare provider is notified, and the process is aborted.

3.2.4 Activity Diagram

An Activity Diagram is used to model the workflow of a particular use case. It

represents the sequence of activities and the flow of control in a system process.
Activity diagrams show how the system and actors interact in each use case scenario.
Below is a high-level activity diagram for the "Submit Patient Data for Analysis" use
case:

42
1. User enters health data → The user inputs their health information into the
system.
 Store health data in database → The system saves the health data for further
processing.
 Decision: Is the user a Health Care Provider?

 Yes:
1. The provider accesses patient data.
2. The system retrieves patient data from the database.
3. The provider views the patient data.
 No:
1. The system sends the health data to a Machine Learning (ML)
model for prediction.
2. The ML model processes the data.
3. The system returns the prediction result.
4. The prediction result is displayed to the user.

Figure 3.1 Activity diagrams

43
3.3Behavioral/Dynamic Modeling

Behavioral/Dynamic Modeling focuses on representing the interactions and behaviors

of the system over time. Unlike static modeling, which shows how a system is
structured, dynamic modeling illustrates how the system evolves and reacts to
different inputs and actions. It focuses on the interactions between system components,
as well as how states change as a result of those interactions. This section covers
Sequence Diagrams and State Diagrams to model the behavior of the Diabetes
Prediction System.

3.3.1 Sequence Diagram

A Sequence Diagram illustrates how objects or components in a system interact over

time to perform a particular function or achieve a goal. It shows the sequence of
messages exchanged between actors (users or external systems) and the system
components involved in a use case or process.

For the Diabetes Prediction System, a sequence diagram can be used to model the
"Predict Diabetes Risk" use case. Below is a high-level description:

Figure 3.2 Sequence Diagram

44
 Actors: Healthcare Provider, Patient, System (Web Application), Machine
Learning Model
Objects: User Interface, Data Processor, ML Model, Database

Sequence Diagram Steps:

1. User enters health data → The input is sent to the Web Application.

2. Web Application stores health data in the database.

3. Web Application sends health data to the ML Model for prediction (if the
user is not a healthcare provider).

4. ML Model processes data and returns prediction result.

5. Web Application displays prediction result to the user.

6. Health Care Provider accesses patient data.

7. Web Application retrieves patient data from the database.

8. Web Application sends patient data to the Health Care Provider.

9. Health Care Provider views patient data.

10. Admin manages system settings.

11. Web Application updates system configurations in the database.

This sequence diagram illustrates the flow of messages between the healthcare
provider, user interface, data processor, machine learning model, and the database to
process and display the diabetes risk prediction.

45
A State Diagram (or Statechart Diagram) is used to model the states of a system or an
object over time, as well as the transitions between those states triggered by events. It
is particularly useful for modeling the lifecycle of an object or entity, showing how it
reacts to various inputs and changes.

For the Diabetes Prediction System, we can model the "Patient Data Processing"
lifecycle, representing the different states of the patient data and how it transitions
through different states until a prediction is made.

Figure 3.3: State Diagram

46
 Steps:

1. Start
 The process begins when the user initiates the system.
 InputData
 The user submits health data into the system.
 PreprocessData
 The system cleans and normalizes the data for further processing.
 ModelPrediction (Nested within PreprocessData)
 The system runs the prediction model on the preprocessed data.
 DisplayResult (Nested within ModelPrediction)
 The prediction result is shown to the user.
 End
 The process completes after displaying the prediction result.

Transitions:

 Waiting for Data → Data Validated: Occurs when the system receives data
from the user and checks it for completeness and correctness.
 Data Validated → Data Processed: Happens after the data is confirmed to be
valid and ready for analysis.
 Data Processed → Prediction Generated: The machine learning model
processes the validated data to make predictions.
 Prediction Generated → Prediction Displayed: The system outputs the
predictions generated by the model.
 Prediction Displayed → Completed State: Final stage where results are
stored, and the process is marked complete.

47
3.4 Class-Based Modeling

Class-based modeling is a technique used in object-oriented design to represent the

structure of the system through classes and the relationships between them. In the
context of the Diabetes Prediction System, class-based modeling helps define the
main components of the system and their interactions. Each class represents a specific
entity or concept in the system, and the relationships between these classes define
how data is managed and processed.

3.4.1 Identifying Classes

In the Diabetes Prediction System, several key entities or objects need to be modeled
as classes. These classes will correspond to the core components of the system and
their attributes. Here are some of the potential classes that can be identified:

Patient: This class represents the patient whose data is used for prediction. It stores
attributes such as age, gender, BMI, blood pressure, glucose level, and family history
of diabetes.

Prediction: This class holds the prediction results generated by the machine learning
model. It represents the outcome of the diabetes risk assessment for a patient.

Figure 3.4 Class-Based Modeling Patient

48
Machine Learning Model: This class handles the machine learning model's
dataprocessing and prediction functionality.

Figure 3.5Class-Based Modeling PredictionDatabase:

This class manages the storage and retrieval of patient and prediction data.

Figure 3.6 Class-Based Modeling Database

49
UserInterface: This class represents the user interface components that allow
healthcare providers and patients to interact with the system.

Figure 3.7 Class-Based Modeling UserInterface:

Validation: This class ensures the validity of input data, such as checking for missing
or incorrect values.

Figure 3.8 Class-Based Modeling Validation:

50
3.4.2 Class Diagram

A Class Diagram visually represents the structure of the system by showing the
classes, their attributes, methods, and the relationships between them. Below is a basic
class diagram for the Diabetes Prediction System.

Figure 3.9 Class Diagram Representation:

51
Explanation:

Relationships:
The Patient class has a one-to-many relationship with the Prediction
class, meaning a patient can have multiple predictions.
The Healthcare Provider interacts with the Patient and Prediction
classes to enter data and generate predictions.
The MachineLearningModel class interacts with the Prediction class
to generate a diabetes risk prediction based on patient data.
The Database class is used by all other classes to store and retrieve
data.
The UserInterface is the entry point for interaction and is connected to
both the Patient and Prediction classes.

52
Summary:

 Identifying Classes: The core entities in the Diabetes Prediction System are
Patient, Healthcare Provider, Prediction, MachineLearningModel,
Database, UserInterface, and Validation.

 Class Diagram: This diagram visually represents the structure of the system,
detailing the attributes and methods of each class and their relationships.

Class-based modeling is crucial for understanding the structure of the system and the
interactions between various components. It also helps guide the design and
implementation of the system by clearly defining how data is stored and processed.

53
Chapter 4: Conclusion and Future Work

4.1 Conclusion

The pre-system aimed to develop a Diabetes Prediction System using machine

learning techniques, providing a web-based platform for early diabetes risk
assessment. The pre-system integrates various machine learning models, including
Logistic Regression, K-Nearest Neighbors (KNN), and Random Forest, to analyze
patient health data and predict the likelihood of diabetes. By leveraging the PIMA
Indian Diabetes Dataset and real-world patient data, the pre-system has demonstrated
the potential to assist healthcare professionals and individuals in identifying diabetes
risks early, thereby facilitating timely medical intervention and lifestyle adjustments.

The pre-system addresses key challenges associated with traditional diabetes

diagnosis methods, such as high costs, time-consuming laboratory tests, and limited
accessibility in remote areas. The user-friendly interface ensures ease of use for both
healthcare professionals and patients, while data encryption and security measures
ensure patient confidentiality and compliance with regulatory standards.

Despite its achievements, the pre-system is not without limitations. The accuracy of
predictions depends on the quality and diversity of training data, and the pre-system
currently focuses primarily on Type 2 Diabetes Mellitus (T2DM). Additionally, real-
time blood sugar monitoring and integration with medical devices are not yet
implemented.

4.2 Future Work

To further enhance the pre-system's capabilities and impact, the following areas of
future work are proposed:
1. Integration with Real-Time Medical Devices

 Future iterations of the pre-system can integrate with wearable health

monitors and glucose sensors to provide real-time diabetes risk
assessment and tracking.

54
2. Enhancement of Machine Learning Models

 Implementing advanced deep learning models, such as convolutional

neural networks (CNNs) and recurrent neural networks (RNNs), could
improve prediction accuracy.
 Incorporating explainable AI (XAI) techniques to provide clear
reasoning behind predictions.
3. Expansion to Other Forms of Diabetes

 Enhancing the pre-system to predict Type 1 Diabetes Mellitus and

Gestational Diabetes by including relevant biomarkers and risk factors.
4. Larger and More Diverse Datasets

 Collaborating with hospitals and research institutions to collect larger

and more diverse patient datasets to improve model generalization and
accuracy.
5. Mobile Application Development

 Developing a mobile application to enhance accessibility, allowing

users to track their health data and receive personalized health
recommendations on the go.
6. Multi-Language Support

 Implementing multi-language functionality to cater to non-English

speaking users and expand the pre-system's usability globally.
7. Integration with Electronic Health Records (EHR)

 Establishing interoperability with existing EHR systems to enable

seamless data exchange between hospitals and the prediction system.
8. Regulatory Compliance and Clinical Validation

 Conducting clinical trials and obtaining regulatory approvals to ensure

the pre-system meets medical industry standards for deployment in
healthcare facilities.By addressing these future improvements, the
Diabetes Prediction Pre-System can become a more robust, accurate,
and widely accessible tool for diabetes prevention and management,
ultimately contributing to better healthcare outcomes worldwide.

55
Appendices

Appendix A: Survey-Based Data

The following table presents a dataset collected from a healthcare survey, containing
anonymized patient health records used for diabetes prediction. The dataset includes
essential attributes such as Patient ID, Age, Gender, BMI, Blood Pressure, Glucose
Level, Cholesterol Level, Physical Activity, Family History, and Diabetes Risk
classification.
The data was gathered through a structured survey and serves as the basis for training
and evaluating the machine learning model for diabetes prediction. Below is a
screenshot of the dataset used in this project:

56
 Appendix B: Additional Test Cases
Table B1: Test Cases for Diabetes Prediction System

Test Case Input Expected Output

Valid Data Submission Age=45, BMI=28, Glucose=140, BP=85 Risk Score: High (85%)

Missing Glucose Value Age=30, BMI=25, Glucose=Null Error: "Glucose value required"

Extreme BMI Age=50, BMI=45, Glucose=160 Risk Score: Very High (95%)

Low-Risk Profile Age=25, BMI=22, Glucose=90 Risk Score: Low (10%)

Appendix C: Diagrams

Figure C1: Use Case Diagram

Description: Illustrates interactions between actors (patients, healthcare providers)
and system functionalities (data input, prediction, reporting).

Figure C2: Entity-Relationship Diagram (ERD)

Description: Shows relationships between database entities (Patient, Prediction,
Doctor) with attributes like Patient ID, Glucose Level, and Risk Score.

57
Appendix D: Dataset Description

PIMA Indian Diabetes Dataset

 Source: UCI Machine Learning Repository

 Samples: 768
 Features: 8 (e.g., Glucose, BMI, Blood Pressure)
 Target Variable: Binary (0 = No Diabetes, 1 = Diabetes)
Summary Statistics:

Feature Mean Std. Deviation

Glucose 120.9 31.97

BMI 31.99 7.88

Blood Pressure 69.1 19.36

Appendix E: User Guide

Steps to Use the Diabetes Prediction System:

1. Registration:

 Visit the web portal and sign up using your email.

 Select your role (Patient/Healthcare Provider).
2. Data Input:

 Enter health metrics: Age, BMI, Glucose, Blood Pressure.

 Click "Submit" to save data.

58
 3. Generate Prediction:

 Click "Predict Risk" to run the machine learning model.

4. View Results:

 Risk score (Low/Medium/High) displayed with recommendations.

 Download report as PDF for medical consultations.

Appendix F: Ethical Considerations

1. Data Anonymization: Patient identifiers removed to comply with

HIPAA/GDPR.

2. Bias Mitigation: Dataset checked for imbalances; SMOTE used to

oversample minority classes.

3. Transparency: Users receive explanations for predictions (e.g., "High glucose

levels contributed to this risk score").

59
References

1. American Diabetes Association. (2023). Diagnosis and Classification of

Diabetes Mellitus. Diabetes Care, 46(Supplement_1), S19–S40.
https://doi.org/10.2337/dc23-S002

2. Smith, J., Patel, R., & Lee, K. (2020). Machine Learning for Diabetes
Prediction: A Comparative Study. Journal of Medical Informatics, 15(3), 45–
60. https://doi.org/10.1016/j.jmi.2020.03.005

3. UCI Machine Learning Repository. (2023). PIMA Indians Diabetes Dataset.

University of California, Irvine. Retrieved from
https://archive.ics.uci.edu/dataset/529/pima+indians+diabetes

4. World Health Organization (WHO). (2023). Global Report on Diabetes.

World Health Organization. Retrieved from
https://www.who.int/publications/i/item/9789241565257

5. Pedregosa, F., et al. (2011). Scikit-learn: Machine Learning in Python.

Journal of Machine Learning Research, 12, 2825–2830.
https://jmlr.org/papers/v12/pedregosa11a.html

60