Chronic Disease Prediction System

Developed a full-stack machine learning web application for early risk assessment of Diabetes and Lung Cancer. The solution integrates a React frontend, Flask REST APIs, and disease-specific machine learning models to deliver real-time predictions from clinical parameters. My primary contribution focused on the complete Diabetes prediction pipeline, including data preprocessing, model evaluation, model selection, and deployment.

Role

ML Developer

Focus Area

Diabetes Prediction

Architecture

React • Flask • ML

Tech Stack

Python • Scikit-learn

Project Objective

Develop a web-based machine learning application capable of providing instant risk assessment for chronic diseases by combining user-friendly interfaces, REST APIs, and trained predictive models into a unified prediction platform.

Business Problem

Early identification of chronic diseases enables timely medical intervention and improved patient outcomes. The objective was to build an accessible web application that processes clinical parameters and generates real-time disease risk predictions for educational and early-awareness purposes.

Solution

Developed separate prediction modules for Diabetes and Lung Cancer.
Integrated a React frontend with Flask REST APIs.
Built disease-specific preprocessing pipelines.
Evaluated multiple machine learning algorithms for each disease.
Serialized trained models for fast real-time inference.
Delivered interactive prediction results through a responsive web interface.

My Contribution (Diabetes Module)

Prepared and preprocessed the Diabetes dataset.
Handled missing values and feature selection.
Applied feature scaling using StandardScaler.
Performed train-test splitting and model evaluation.
Compared Logistic Regression, Decision Tree, Random Forest, XGBoost, and Voting Classifier.
Selected Decision Tree Classifier based on evaluation metrics.
Integrated the trained model into the Flask prediction pipeline.

Model Performance

Selected Model: Decision Tree Classifier
Accuracy: Approximately 82.5%
Precision: Approximately 70.8%
Recall: Approximately 72.3%
ROC-AUC Score: 0.8390

Application Workflow

User enters clinical parameters through the React interface.
Input data is sent to the Flask backend using REST APIs.
The preprocessing pipeline validates and transforms the input.
The trained machine learning model generates a prediction.
The prediction result is returned and displayed to the user.

Engineering Highlights

Implemented a modular architecture supporting multiple disease prediction models.
Built separate preprocessing pipelines for different datasets.
Integrated machine learning models with Flask REST APIs.
Serialized trained models for efficient inference.
Created a responsive React-based user interface.
Evaluated multiple classification algorithms before final model selection.

Business Value

Provides instant risk assessment through an intuitive web application.
Demonstrates practical integration of machine learning with full-stack development.
Supports educational awareness through accessible predictive analytics.
Establishes a scalable foundation for adding additional disease prediction modules.

Skills Demonstrated

Python • Machine Learning • Scikit-learn • Decision Tree • Model Evaluation • Data Preprocessing • Feature Engineering • StandardScaler • Flask • REST APIs • React • Pandas • NumPy • XGBoost