Blood Donation Analysis

This project predicts whether a person will donate blood based on donation history features. The repository includes:

• train.py to train the model and save the artifacts
• app.py to load the saved model and expose a Flask prediction API

Setup

Install the required packages before running any script:

pip install pandas numpy matplotlib seaborn scikit-learn imbalanced-learn xgboost scikit-optimize flask flask-cors joblib

How to Run

1. Train the model

Run the training script from the project root:

python train.py

This will:

• load blood-train.csv
• perform preprocessing, feature engineering, SMOTE balancing, and XGBoost tuning
• save the trained model to blood_donation_xgb_model.joblib
• save the scaler to blood_donation_scaler.joblib

2. Start the prediction API

After training completes, run:

python app.py

The Flask server starts on http://127.0.0.1:5000.

3. Call the prediction endpoint

Send a POST request to /predict with JSON input:

curl -X POST http://127.0.0.1:5000/predict ^
  -H "Content-Type: application/json" ^
  -d "{\"months_last\":10,\"num_donations\":5,\"total_volume\":1250,\"months_first\":36}"

Example response:

{
  "prediction_text": "Will Donate",
  "confidence": 0.85
}

Model Performance

Model	Accuracy	Precision	Recall	F1-Score
Logistic Regression	0.7625	0.8234	0.5463	0.6667
Decision Tree	0.7891	0.8148	0.6082	0.6982
Random Forest	0.8061	0.8378	0.6642	0.7417
Support Vector Classifier	0.7823	0.8095	0.5864	0.6789
XGBoost (Optimized)	0.8300	0.8547	0.8696	0.8620

Best Model

XGBoost (Optimized) performed best across all four metrics and is the model saved by the training pipeline.

Notes

• Run the scripts from the project root so the CSV and model paths resolve correctly.
• If you retrain the model, the .joblib files will be overwritten with the new artifacts.

UI Screenshots

The application provides an interactive web interface for blood donation predictions. Below are screenshots showcasing the prediction interface:

Prediction Example 1: "Will Not Donate"

Input: Months since last donation: 12, Number of donations: 23, Total volume donated: 12500 cc, Months since first donation: 9 Result: Will Not Donate

Prediction Example 2: "Will Donate"

Input: Months since last donation: 2, Number of donations: 20, Total volume donated: 5000 cc, Months since first donation: 45 Result: Will Donate

Prediction Example 3: "Will Donate" (High Frequency Donor)

Input: Months since last donation: 2, Number of donations: 50, Total volume donated: 12500 cc, Months since first donation: 98 Result: Will Donate