考虑社会关系影响差异和动态性的社会化推荐

Social recommendation based on considerations of different effects and dynamicity of social relations

随着社交媒体的发展,社交活动日益普及并产生丰富的社会关系。融合社会关系进行推荐可以缓解传统推荐系统面临的数据稀疏和冷启动问题。已有很多成功利用社会关系和评分信息进行推荐的算法,然而它们没有充分地挖掘不同的社会关系对用户的不同影响以及在不同时间段中社会关系的不同影响,这导致推荐效果的下降。基于对现实中社会关系影响复杂性的充分考虑,提出了新的考虑社会关系影响差异和动态性的社会化推荐算法。在Epinions数据集上的实验结果表明该方法可以提高推荐质量。复杂度分析也表明该方法具有可扩展性,能够适应大数据集的要求。

With the development of social media,social activities are more and more popular among people and rich social relations have been produced.Fusing social relations to recommend can potentially alleviate data sparse and cold start problems that traditional recommender systems face.Many successful recommendation algorithms employing both social relations and rating information have been proposed in recent years.However,these existing algorithms do not consider much on the differences of the influences generated by each social relation and the temporal changes of social factors,which degraded the performance of recommendation.Based on the fact that the social relations in the real world are very complex,this paper proposed new social recommendation algorithms that considered the different effects and dynamicity of social relations.The experimental results on Epinions dataset show that this method can improve the recommendation quality.The complexity analysis also indicates that this approach can be applied to large datasets.