Diabetes Prediction Web Application

A full-stack web application for diabetes prediction using a machine learning ensemble model. The system combines seven different algorithms in a soft voting ensemble to provide accurate predictions with confidence scores.

Overview

This application uses state-of-the-art machine learning techniques to predict diabetes based on diagnostic measurements. The model achieves 83% accuracy on test data with a soft voting ensemble approach.

Features

• Machine Learning Model: Soft voting ensemble combining 7 algorithms (KNN, SVM, Decision Tree, Random Forest, Logistic Regression, XGBoost, LightGBM)
• REST API: Flask-based backend with JSON endpoints
• Interactive Web Interface: Responsive design for easy data entry and visualization
• Real-time Predictions: Instant results with probability distributions
• Confidence Scoring: Detailed breakdown of prediction confidence

Model Performance

• Test Accuracy: 83.00%
• Sensitivity (Recall): 92.00%
• Specificity: 74.00%
• ROC-AUC Score: 0.8812
• PR-AUC Score: 0.8772

Project Structure

diabetes_web_app/
├── app.py                      # Flask application server
├── train_model.py              # Model training script
├── requirements.txt            # Python dependencies
├── README.md                   # This file
├── models/                     # Trained model files (generated)
│   ├── diabetes_model.pkl
│   ├── scaler.pkl
│   └── feature_names.pkl
├── templates/
│   └── index.html              # Web interface
└── static/
    ├── css/
    │   └── style.css           # Stylesheet
    └── js/
        └── app.js              # Frontend JavaScript

Installation

Prerequisites

• Python 3.8 or higher
• pip package manager

Setup Steps

1. Navigate to the project directory

   cd diabetes_web_app

2. Install dependencies

   pip install -r requirements.txt

   Or install packages individually:

   pip install flask numpy pandas scikit-learn xgboost lightgbm imbalanced-learn

3. Train the model

   Ensure diabetes.csv is in the parent directory, then run:

   python train_model.py

   This will create the model files in the models/ directory.

4. Start the application

   python app.py

5. Access the application

   Open your browser and navigate to:

   http://localhost:5000
   

Quick Start

Windows

Double-click run.bat to automatically train the model (if needed) and start the server.

Mac/Linux

chmod +x run.sh
./run.sh

Usage

Web Interface

1. Open http://localhost:5000 in your browser
2. Enter patient information in the form (8 required fields)
3. Click "Predict" to get the prediction
4. View results including:
   • Prediction label (Diabetic/Non-Diabetic)
   • Confidence percentage
   • Probability distribution

API Endpoints

Health Check

GET /health

Response:

{
  "status": "healthy",
  "model_loaded": true,
  "scaler_loaded": true
}

Get Features

GET /features

Response:

{
  "features": ["Pregnancies", "Glucose", "BloodPressure", "SkinThickness", "Insulin", "BMI", "DiabetesPedigreeFunction", "Age"],
  "count": 8
}

Make Prediction

POST /predict
Content-Type: application/json

Request body:

{
  "pregnancies": 6,
  "glucose": 148,
  "bloodPressure": 72,
  "skinThickness": 35,
  "insulin": 0,
  "bmi": 33.6,
  "diabetesPedigreeFunction": 0.627,
  "age": 50
}

Response:

{
  "prediction": 1,
  "prediction_label": "Diabetic",
  "probability": {
    "non_diabetic": 0.23,
    "diabetic": 0.77
  },
  "confidence": 77.0,
  "input_features": {
    "Pregnancies": 6,
    "Glucose": 148,
    "BloodPressure": 72,
    "SkinThickness": 35,
    "Insulin": 0,
    "BMI": 33.6,
    "DiabetesPedigreeFunction": 0.627,
    "Age": 50
  }
}

Input Features

Feature	Description	Range	Unit
Pregnancies	Number of times pregnant	0-17	count
Glucose	Plasma glucose concentration	0-199	mg/dL
BloodPressure	Diastolic blood pressure	0-122	mm Hg
SkinThickness	Triceps skin fold thickness	0-99	mm
Insulin	2-Hour serum insulin	0-846	μU/mL
BMI	Body Mass Index	0-67.1	kg/m²
DiabetesPedigreeFunction	Diabetes heredity function	0.078-2.42	score
Age	Age in years	21-81	years

Model Architecture

Ensemble Approach

The model uses a soft voting ensemble that combines predictions from seven different algorithms:

1. K-Nearest Neighbors (KNN) - k=3
2. Support Vector Machine (SVM) - RBF kernel, gamma=2, C=1
3. Decision Tree - max_depth=5
4. Random Forest - 10 estimators
5. Logistic Regression - max_iter=1000
6. XGBoost - 100 estimators, learning_rate=0.1
7. LightGBM - 100 estimators, learning_rate=0.1

Data Preprocessing

1. Class Balancing: RandomOverSampler to handle imbalanced dataset (500:500 ratio)
2. Feature Scaling: MinMaxScaler to normalize all features to [0, 1] range
3. Train-Test Split: 80% training, 20% testing with stratification

API Usage Examples

Python

import requests

url = "http://localhost:5000/predict"
data = {
    "pregnancies": 6,
    "glucose": 148,
    "bloodPressure": 72,
    "skinThickness": 35,
    "insulin": 0,
    "bmi": 33.6,
    "diabetesPedigreeFunction": 0.627,
    "age": 50
}

response = requests.post(url, json=data)
result = response.json()

print(f"Prediction: {result['prediction_label']}")
print(f"Confidence: {result['confidence']:.2f}%")

cURL

curl -X POST http://localhost:5000/predict \
  -H "Content-Type: application/json" \
  -d '{
    "pregnancies": 6,
    "glucose": 148,
    "bloodPressure": 72,
    "skinThickness": 35,
    "insulin": 0,
    "bmi": 33.6,
    "diabetesPedigreeFunction": 0.627,
    "age": 50
  }'

JavaScript

const data = {
  pregnancies: 6,
  glucose: 148,
  bloodPressure: 72,
  skinThickness: 35,
  insulin: 0,
  bmi: 33.6,
  diabetesPedigreeFunction: 0.627,
  age: 50
};

fetch('http://localhost:5000/predict', {
  method: 'POST',
  headers: {
    'Content-Type': 'application/json',
  },
  body: JSON.stringify(data)
})
.then(response => response.json())
.then(result => console.log(result));

Test Cases

Non-Diabetic Patient

Pregnancies: 1
Glucose: 85
Blood Pressure: 66
Skin Thickness: 29
Insulin: 0
BMI: 26.6
Diabetes Pedigree Function: 0.351
Age: 31

Expected Result: Non-Diabetic

Diabetic Patient

Pregnancies: 6
Glucose: 148
Blood Pressure: 72
Skin Thickness: 35
Insulin: 0
BMI: 33.6
Diabetes Pedigree Function: 0.627
Age: 50

Expected Result: Diabetic

Troubleshooting

Module Not Found Error

Ensure all dependencies are installed:

pip install -r requirements.txt

Dataset Not Found Error

Verify that diabetes.csv is in the parent directory:

Project/
├── diabetes.csv          (must be here)
└── diabetes_web_app/
    └── train_model.py

Port Already in Use

Change the port in app.py:

app.run(debug=True, host='0.0.0.0', port=5001)

Model Files Not Found

Run the training script first:

python train_model.py

Deployment

Local Development

python app.py

Runs on http://localhost:5000 with debug mode enabled.

Production with Gunicorn

pip install gunicorn
gunicorn -w 4 -b 0.0.0.0:5000 app:app

Docker

FROM python:3.9-slim
WORKDIR /app
COPY requirements.txt .
RUN pip install -r requirements.txt
COPY . .
EXPOSE 5000
CMD ["python", "app.py"]

Build and run:

docker build -t diabetes-app .
docker run -p 5000:5000 diabetes-app

Technical Stack

Backend

• Flask 3.0+
• Python 3.8+
• scikit-learn
• XGBoost
• LightGBM
• NumPy
• Pandas
• imbalanced-learn
• joblib

Frontend

• HTML5
• CSS3
• JavaScript (ES6+)

Important Disclaimer

This application is intended for educational and research purposes only. It should not be used as a substitute for professional medical advice, diagnosis, or treatment. Always consult with qualified healthcare professionals for medical decisions.

License

This project is available for educational purposes.

Acknowledgments

• Dataset: Pima Indians Diabetes Database
• Libraries: scikit-learn, XGBoost, LightGBM, Flask