A Machine Learning-Based Medical System for Accurate Blood Donor–Recipient Compatibility Prediction

Abstract

Blood transfusion is an essential medical process that saves millions of lives annually, yet its success critically depends on accurate donor–recipient compatibility assessment. Conventional blood matching systems predominantly rely on ABO and RhD blood group typing, neglecting extended antigen profiles and patient transfusion histories. This limitation elevates the risk of transfusion reactions, particularly in patients requiring repeated transfusions, such as those with thalassemia, sickle cell disease, or cancer. This paper presents a browser-based medical system that employs machine learning algorithms—Random Forest and Naïve Bayes—to predict donor–recipient compatibility using key clinical parameters including ABO/RhD blood group, crossmatch test results, transfusion history, disease type, and extended antigen data. The system supports role-based hospital workflows for administrators, receptionists, laboratory technicians, and patients. Developed with Python, Flask, MySQL, scikit-learn, and Bootstrap, the proposed system demonstrates improved prediction accuracy, reduced manual errors, and enhanced transfusion safety over traditional methods.