Efficient routing on finding recommenders for trust-aware recommender systems

Abstract

The trust-aware recommender system (TARS) is a newly proposed trust-aware application. It is able to solve the data sparseness problem of the conventional recommender systems. One of the basic research challenges in TARS is to find the recommenders efficiently. Existing works of TARS use the strategy of random walk to find the recommenders, which is obviously low efficiency. Though the trust network has been verified to be the scale-free network, due to the small power of its degree distributions, we have verified via experiments that the prediction coverage of TARS is very limited by applying the classical routing protocol of scale-free networks directly. We therefore propose a routing protocol for TARS, which is able to efficiently find reliable recommenders for the users of TARS. Our protocol is able to achieve much higher prediction coverage than the classical routing protocol of scale-free networks, while the computational complexity is greatly reduced comparing with existing works of TARS.