Predictive modelling for patient care using federated learning systems combined with a private blockchain data environment and using ranked set sampling

Abstract Recent advancements in information technology have enhanced patient care, yet data breaches in healthcare still pose major privacy and security challenges. This study introduces a predictive model designed to safeguard sensitive medical data while maintaining high predictive accuracy by integrating blockchain, deep learning, federated learning (FL), and ranked set sampling (RSS). The model employs a long short-term memory network within a FL framework and incorporates a private blockchain for secure, decentralized, and tamper-resistant data management. To improve efficiency, RSS is used to prioritize the most informative samples before model training. The model was first tested on a breast cancer dataset, achieving a test loss of 0.29, an accuracy of 93.75%, and an AUC of 0.94, showing excellent classification performance. Further validation on a Hodgkin Lymphoma dataset, characterized by class imbalance, yielded 85% accuracy and an AUC of 0.91, confirming strong discriminative ability. The model was also validated using a lung cancer dataset. These results highlight the model’s robustness in handling heterogeneous, imbalanced clinical data. By combining FL, blockchain, and RSS, the proposed system ensures data privacy, computational efficiency, and scalability, making it a reliable and secure solution for modern healthcare applications.