This paper introduces a web-based application that predicts a person's ABO-Rh blood group from fingerprint images using machine learning. The system provides a non-invasive, fast, and accessible solution for estimating blood groups, especially useful in emergency or low-resource settings. The application uses Histogram of Oriented Gradients (HOG) to extract key features from fingerprint images. These features are then analyzed using a Support Vector Machine (SVM) classifier trained on a labeled dataset of fingerprint samples across different blood groups. The model uses class balancing techniques to improve prediction fairness across all classes. Users upload fingerprint images through a simple web interface, and the system provides real-time predictions with confidence scores and probability breakdowns. It also allows users to correct or confirm predictions, which helps improve the model over time by saving corrected samples to the dataset.
Fingerprint, Blood Group, SVM, HOG, Machine Learning