A DIABETES DIAGNOSTIC SYSTEM USING MACHINE

LEARNING

Capstone Project-1 (CSE 339)

DECLARATION:
I declare that this assignment is teamwork. We have not copied it from any
others’ work or any other source except where due acknowledgement is made
explicitly in the text, nor has any part been written for me by any other
person.

SUBMITTED BY:
SUBMITTED
ADITYA SAH (12111525)
TO:
APSARA R (12108076)
Amarinder
ANIRBAN MISHRA Kaur
(12106644)
21482
ASHIN VINCENT
 INTRODUCTION

The presence of diabetes is steadily rising to popularity as a global health concern. Type 1 diabetes and Type 2 diabetes are the
various categories into which it falls. The chance of developing problems such as heart disease, diabetic nerve damage, or renal
failure can be reduced with earlier identification. However, certain recognized diagnostic techniques can be time-consuming
and intrusive. When medical data is properly classified and patterns are recognized, machine learning (ML) algorithms may be
utilized effectively to diagnose and forecast diabetes. With a focus on providing significant algorithms, data sets, and
performance metrics, this paper reviews recent advancements in the use of machine learning (ML) systems for diabetes
diagnosis. Furthermore, prospects and problems in the field of using machine learning to diabetes prediction are discussed.

Keywords—logistic regression, random forest, linear SVM, machine learning models, prediction

However, Gabor and others can forecast the likelihood of developing diabetes by analyzing large amounts of data using
machine learning. Some of the most popular machine learning algorithms for diabetes are covered in this article, along with
how they are applied for diabetes diagnosis and the challenges associated with their effective usage. Diabetes is characterized
by the aid of continual hyperglycemia because of insulin deficiency or resistance. Diabetes results in excessive complications
which include cardiovascular ailment, kidney failure, nerve harm, and blindness if left undiagnosed or poorly managed.
Traditional diagnostic methods for diabetes depend on biochemical assessments which include fasting blood glucose, oral
glucose tolerance exams (OGTT), glycated hemoglobin (HbA1c) measurements, etc. While these diagnostic assessments are
effective, they often require a specialized system, educated employees, and repeated medical institution visits, making them
less handy in below resource or rural regions. Additionally, the reliance on static thresholds for diagnosis might not account for
man or woman variability, together with age, ethnicity which could affect diabetes risk. When deployed on cloud-based
structures, rule-primarily based systems gain extra abilities, such as scalability, remote get right of entry, and seamless
integration with digital health information. Cloud infrastructures enable real-time statistics processing, allowing healthcare
providers to deliver timely diagnoses and screen patient progress no matter geographical constraints. Moreover, cloud systems
can help privateness sensitive facts management protocols, ensuring compliance with policies like Health Insurance Portability
and Accountability Act (HIPAA) and General Data Protection Regulation (GDPR) at the same time as fostering patient
acceptance as true within digital fitness answers. This review paper explores the potential of hybrid systems that combine rule-
primarily based diagnostic fashions with cloud based infrastructures in revolutionizing diabetes analysis. It examines the
modern techniques of such systems, their underlying methodologies, and real-global packages even as evaluating their
strengths, limitations, and moral considerations. Furthermore, the paper highlights rising trends and opportunities for
innovation, which include the combination of artificial intelligence, system learning, and personalized medicine into those
structures. By synthesizing the to-be-had evidence, this assessment goals to make contributions to the growing frame of know-
how on leveraging generation to create handy, dependable, and affected person-centered diagnostic solutions for diabetes care.
 DETAILS OF PAPER READ

Machine learning for diabetes clinical decision support: a review

DOI: [Link]

Diabetes detection based on machine learning and deep learning approaches

DOI: [Link]

Diabetes Prediction using Machine Learning Algorithms

DOI: [Link]

WORK ROLE DISTRIBUTION

A. USER INTERFACE (UI) DESIGN (ADITYA SAH)

Tasks:
 Design the layout for the web or mobile app (e.g., login, prediction input form, results display).
 Select color schemes, typography, and themes for a user-friendly experience.
 Create wireframes and prototypes using tools like Figma, Adobe XD, or Sketch.
 Responsibility: UI/UX Designer or Front-End Developer.

1. Form and Input Fields for User Data

Tasks:
Develop input forms for user data collection (e.g., age, glucose level, BMI).
Implement input validation to ensure data integrity (e.g., numeric only, ranges).
Provide dropdowns, sliders, or other interactive components for data entry.
Responsibility: Front-End Developer.

B. MACHINE LEARNING MODEL INTEGRATION (ANIRBAN MISHRA)

Tasks:
Load and serve the pre-trained machine learning model (e.g., trained with algorithms like Logistic Regression, Random
Forest, or Neural Networks).
Use frameworks to expose the model as an API.
Responsibility: Data Scientist and Back-End Developer.

1. API Development

Tasks:

Design and implement RESTful or GraphQL APIs for communication between the front-end and back-end.

Create endpoints for:

Submitting user data for prediction.

Retrieving prediction results.

 C. ALGORITHM HANDLING (APSARA R.)
Data Collection and Preparation
 Tasks:
 Collect relevant datasets (e.g., Pima Indians Diabetes Dataset, or custom datasets).
 Preprocess the data:
 Handle missing values (e.g., imputation or removal).
 Normalize or standardize numerical features.
 Encode categorical variables (e.g., one-hot or label encoding).
 Feature scaling for ML algorithms that require it (e.g., SVM).
 Perform exploratory data analysis (EDA) to identify trends and correlations.
 Responsibility: Data Engineer and Data Scientist.

. Feature Engineering

 Tasks:
 Identify key features influencing diabetes prediction (e.g., age, BMI, glucose levels).
 Engineer new features based on domain knowledge (e.g., age groups, glucose-BMI interactions).
 Perform feature selection using techniques like correlation matrices, mutual information, or recursive feature
elimination (RFE).
 Responsibility: Data Scientist.

D. CLOUD MODEL DEPLOYMENT (ASHIN VINCENT)

Cloud Platform Setup
 Tasks:
 Choose a cloud provider (e.g., AWS, Google Cloud Platform (GCP), Microsoft Azure).
 Set up cloud services for:
 Computing (e.g., AWS EC2, GCP Compute Engine, Azure Virtual Machines).
 Storage (e.g., AWS S3, GCP Cloud Storage, Azure Blob Storage) for model files or user data.
 Databases (e.g., RDS, Cloud SQL, or NoSQL databases like DynamoDB or Firestore).
 Configure Virtual Private Cloud (VPC) for secure network access.
 Responsibility: Cloud Engineer.

Containerization
 Tasks:
 Package the machine learning model and its dependencies into a container using Docker.
 Define a Dockerfile that includes:
 The base image (e.g., Python or TensorFlow runtime).
 Required dependencies (e.g., Flask, FastAPI, Scikit-learn).
 Code for serving the model.
 Test the container locally before pushing to the cloud.
 Responsibility: ML Engineer or DevOps Engineer.