Machine Learning Based Blood Group Prediction on Fingerprint Images

Abstract – This paper provides a methodology for predicting blood-type classification which uses fingerprint images as input with the help of deep learning to propose an effective technique that is non-invasive, easy and more validated than conventional techniques. The main goal of this work is to determine blood group via fingerprint patterns by utilizing CNN and VGG16 architectures. The dataset of labelled fingerprint images was preprocessed before then being utilized to train and also for validate the models and we evaluated it through precision-recall curves, confusion matrices and class-wise accuracy analysis. The results indicate that VGG16 achieves higher accuracy is most of the blood group categories, while the CNN model shows a marginally better performance in classifying the A+ group. Both architectures deliver reliable outcomes with AUC scores exceeding 0.93 for all classes. These findings confirm that predicting blood groups from fingerprint patterns is not only practical but also effective with promising applications in healthcare, forensic science and biometric security. Key Words: BloodGroup, Fingerprint, VGG16 Framework, CNN, biometric Authentication.