Many-to-One Stable Matching for Prediction in Social Networks
Abstract: Stable matching investigates how to pair elements of two disjoint sets with the purpose to achieve a matching that satisfies all participants based on their preference lists. In this paper, we consider the case of matching with incomplete information in a social network where agents are not fully connected. A new many-to-one matching algorithm is proposed based on the classical Gale-Shapley algorithm with constraints of given network topology. In simulated experiments, we find that the matching outcomes in scale-free networks yield the best average utility with least connective costs compared to other structured networks in one-to-one problems. But in many-to-one matching cases, network structure has no significant influence on matching utilities. We also apply the new matching model to a real-world social network matching problem and we find a significant increase of accuracy in matching pair prediction comparing to classical methods.
Loading