Abstract:Recommendation system is becoming one of the most important components in many e-commerce websites and web applications.In the past decades,collaborative filtering (CF) is the most success approach to handle the recommendation task,which leverage users' historical rating data to model users and items preference.As an early approach,the CF algorithm is seriously affected by the data spare problem,which means that most users rate only a few items.This problem distinctly degenerates the performance of CF-based methods.To address this problem,introducing auxiliary information for CF algorithm becomes a necessary trend to improve recommendation accuracy.Therefore,with the development of social media,social recommendation algorithms have been proposed and presented as a potential way to solve this problem.These methods utilizing users' social information to help modeling user preference for recommendation.However,most of them simply make use of the social network information with binary format directly,and thus ignore the fact that the trust degrees between each pair of users are full of variety.In this paper,we propose a new method to measure trust degrees,i.e.,how much degrees will users trust in other users,and to integrate these degrees for social recommendation.Advancing previous works,in additional of rating similarity,we focus on the social information to estimate trust degrees and propose the implicit similarity in trust.Specially,we take the indirect social influence of users into account to estimate trust degrees between each pair of users.We obtain the social preferences which contains indirect influence by factorizing the social matrix.Based on these preferences,we propose the implicit similarity in trust for users,which exactly reflect social relations.In addition,based on the intuition that a user plays different preferences with role of truster and trustee,our proposed implicit similarity in trust is consist of two parts:users as truster and as trustee.Moreover,on account of the fact that both rating and social data are very spare,we take both rating and social data into account to enhance the evaluation of trust degrees.In this way,after synthesizing both rating similarity and implicit similarity in trust for each pair of users,we establish an accurate social recommendation model.Furthermore,instead of directly estimating the trust degrees,we propose an adaptive similarity learning model based on both rating similarity and implicit similarity in trust to evaluate trust degrees for recommender system.We conduct several experiments on both Epinions and Ciao datasets,and compared our model with several state-of-the-art algorithms.The experiments demonstrate that the proposed approach fully exploits implicit information in rating data and social network,and efficiently improves the performance of social recommendation.Specially,the results are evaluated in terms of two error based metrics (MAE and RMSE) and rank-sensitive metrics (precision,recall and NDCG),and indicate that our approach is robust in both tasks of rating prediction and top-n recommendation tasks.We also demonstrate the results in case of sparse rating data or spare social data,which outperforms previous works.In summary,our approach,fully exploiting implicit information in rating and social data,can efficiently improve recommendation accuracy.

Social Recommendation Based on Implicit Friends Discovering Via Meta-Path

Adaptive Implicit Friends Identification over Heterogeneous Network for Social Recommendation.

Exploring Implicit Relationships in Social Network for Recommendation Systems

A Novel Social Recommendation Method Fusing User’s Social Status and Homophily Based on Matrix Factorization Techniques

Extract Implicit Semantic Friends and Their Influences from Bipartite Network for Social Recommendation

Social Recommendation With Learning Personal and Social Latent Factors

A recommender system fused with implicit social information through network representation learning

Fusion of Personalized Implicit Relations for Social Recommendation

JUST-BPR: Identify Implicit Friends with Jump and Stay for Social Recommendation.

Recommender System Using Implicit Social Information.

Revealing and incorporating implicit communities to improve recommender systems.

Incorporating Metapath Interaction on Heterogeneous Information Network for Social Recommendation

Social Recommendation Algorithm Fusing User Interest Social Network

Social Recommendation Algorithm Using Implicit Similarity in Trust

Learning to Recommend with Social Contextual Information from Implicit Feedback

Social Recommendation with Self-Supervised Metagraph Informax Network

Distinguishing Social Ties In Recommender Systems By Graph-Based Algorithms

Exploiting Implicit Item Relationships For Recommender Systems

SimilarMF - A Social Recommender System Using an Embedding Method.

Personalized recommendation based on implicit social network of researchers

Leveraging Implicit Relations for Recommender Systems