CKD Detection System

A web-based application for predicting Chronic Kidney Disease (CKD) using a machine learning model. This project includes data preprocessing, model training, and a user-friendly interface for real-time predictions.

Features

• CKD Prediction: Predicts the risk of CKD based on user-provided medical data.
• User-Friendly Interface: A simple web interface for entering patient data.
• PDF Report Generation: Generates a downloadable PDF report of the prediction results.
• Fallback Prediction: Includes a rule-based fallback mechanism for predictions if the machine learning model is unavailable.

Tech Stack

• Backend: Flask
• Frontend: HTML, CSS, JavaScript
• Machine Learning: scikit-learn, pandas, numpy
• PDF Generation: ReportLab

Project Structure

.
├── app.py                      # Main Flask application
├── requirements.txt            # Project dependencies
├── dataset/
│   └── final.csv               # Raw dataset
├── final/
│   ├── ckd_model.pkl           # Trained machine learning model
│   └── label_encoders.pkl      # Saved label encoders
├── src/
│   ├── preprocess.py           # Data preprocessing script
│   ├── train.py                # Model training script
│   └── evaluation.py           # Model evaluation script
├── static/
│   ├── css/
│   └── js/
└── templates/
    └── index.html              # HTML template for the web interface

Setup and Installation

1. Clone the repository:

   git clone https://github.com/your-username/ckd-detection-system.git
   cd ckd-detection-system

2. Create and activate a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows, use `venv\Scripts\activate`

3. Install the dependencies:

   pip install -r requirements.txt

4. Run the preprocessing and training scripts (optional, as the trained model is provided):

   python src/preprocess.py
   python src/train.py

5. Run the Flask application:

   python app.py

6. Open your web browser and navigate to http://127.0.0.1:5000/

Usage

1. Enter the patient's medical data in the web interface.
2. Click the "Predict" button to get the CKD risk prediction.
3. Download the PDF report for a detailed summary of the results.

Model Details

• Algorithm: Random Forest Classifier
• Features: Serum Creatinine, Hemoglobin, Albumin, Specific Gravity, Packed Cell Volume, Red Blood Cell Count, Diabetes Mellitus, Hypertension
• Training Data: preprocessed_final_ckd.csv

Scripts

• src/preprocess.py: This script loads the raw data from dataset/final.csv, performs label encoding on categorical features, and saves the processed data and encoders.
• src/train.py: This script trains a Random Forest model on the preprocessed data and saves the trained model to final/ckd_model.pkl.
• src/evaluation.py: This script evaluates the trained model using various metrics and generates a classification report and confusion matrix.