Abstract
A computer-implemented platform matches living organ donors with recipients using self-reported profile data and social-network data. A request is routed by organ type to kidney, liver, or bone marrow logic. A Friend2Vec graph embedding model constructs a weighted social graph, performs biased random walks, and trains a Skip-Gram network to produce user embeddings. Friend2Vec adds a donation-aware bias parameter that preferentially traverses edges associated with successful donation events. Embeddings are fused with medical and geographic attribute vectors using a learned fusion layer with layer normalization. A multi-layer perceptron predicts P(Donation | Donor, Recipient) from concatenated donor and recipient embeddings, their element-wise difference, and product. Organ-specific heuristic filters apply ABO and proxies such as body size for liver, or age and ethnicity-based HLA likelihood for bone marrow. Remaining candidates are ranked by a weighted composite score. For kidney paired exchange, an ILP selects exchange cycles with edge weights based on predicted follow-through.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Anonymous, "Multi-Organ Living Donor Matching System Using Social Graph Embedding with Donation-Aware Bias and Heuristic Medical Compatibility Scoring", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/10699