Blood Donation Prediction using ML

Blood Donation Prediction Using Machine Learning

Linkedin: https://www.linkedin.com/posts/harishwaran-vl_machinelearning-ai-datascience-activity-7300641899906768896-pBF0?utm_source=share&utm_medium=member_desktop&rcm=ACoAADsBg0IBEc3FDwTSODKTv75sSLL41wgAAiQ

Problem Statement

Blood donation is a vital aspect of public healthcare, yet identifying and engaging potential donors remains a challenge. Traditional outreach methods are often inefficient and not targeted, resulting in low donor retention and limited availability during critical periods.

To address this issue, a predictive model was needed to help blood banks and donation centers anticipate donor behavior and improve their donor engagement strategies.

Project Objective

The objective of this project was to build a machine learning model that predicts whether an individual is likely to donate blood in the near future based on their donation history. This model would support healthcare organizations in proactively reaching out to potential donors, ensuring timely blood supply and reducing the burden on manual follow-ups.

Additionally, the project included developing a web-based application using Flask, deployed on an AWS EC2 instance, to make the prediction system accessible and user-friendly.

Dataset Overview

• Source: UCI Machine Learning Repository - Blood Transfusion Service Center dataset

• Features:

  • Recency: Months since the last donation

  • Frequency: Total number of donations

  • Monetary: Total blood donated (in c.c.)

  • Time: Months since first donation

• Target Variable: Whether the donor donated blood in the last campaign (Yes/No)

Technical Workflow

1. Data Preprocessing

• Checked for missing values and cleaned the dataset

• Normalized numerical features for better model performance

• Split the data into training and testing sets (80/20)

2. Model Selection and Training

• Tried various classification algorithms:

  • Logistic Regression (baseline)

  • Random Forest Classifier

  • Support Vector Machine

  • K-Nearest Neighbors

• Used GridSearchCV to tune hyperparameters and improve model accuracy

• Evaluated using metrics like:

  • Accuracy

  • Precision

  • Recall

  • F1-score

  • ROC-AUC Curve

3. Model Deployment

• Built a lightweight Flask web application to serve the prediction model

• The UI accepts user input (Recency, Frequency, Monetary, Time) and displays whether the user is likely to donate

• Hosted the Flask app on an AWS EC2 instance, making it accessible over the internet

Results and Performance

• Achieved an accuracy of over 90% using Random Forest and SVM

• The model showed strong recall, which is crucial for not missing out on potential donors

• End-users could obtain predictions instantly through a clean and intuitive web interface

Web App Features

• User input form to enter donation history

• Real-time prediction of donation likelihood

• AWS EC2 hosting for cloud accessibility

• Simple, responsive UI with clear output messaging

Key Achievements

• Delivered a complete ML pipeline from data preprocessing to model deployment

• Gained hands-on experience with:

  • Supervised classification algorithms

  • Flask web development

  • AWS EC2 deployment

• Created a solution with potential real-world application in the healthcare and nonprofit sectors

Impact and Applications

• Helps blood banks and donation centers forecast donation likelihood

• Supports targeted donor outreach, improving campaign effectiveness

• Encourages data-driven donor engagement strategies

Technologies Used

• Python: Pandas, NumPy, Scikit-learn, Matplotlib

• Machine Learning: Logistic Regression, Random Forest, SVM, KNN

• Web Framework: Flask

• Cloud: AWS EC2 for deployment

• DevOps: GitHub for version control