Recommending with limited number of trusted users in social networks
Weiwei Yuan, Donghai Guan, Asad Masood Khattak · 2018
To estimate the reliability of an unknown node in social networks, existing works involve as many opinions from other nodes as possible. Though this makes it possible to approximate the real property of the unknown nodes, the computational complexity increases as the scale of social networks getting bigger and bigger. We therefore propose a novel method which involve only limited number of social relations to predict the trustworthiness of the unknown nodes. The proposed method involves four rating prediction mechanisms: FM use the recommendation given by the most reliable recommender with the shortest trust propagation distance from the active user as the predicted rating, FMW weights the recommendation in FM, FA uses the mean value of recommendations with the shortest trust propagation distance from the active user as the predicted rating, and FAW weights recommendations in FA. The simulation results show that the proposed method can greatly reduce the rating prediction calculation, while the rating prediction losses are reasonable.