一种基于自外积异构信息网络的改进推荐模型

An Improved Recommendation Model with Self-outer Product Enhanced Heterogeneous Information Network

网络嵌入是近年来大数据领域的研究热点。异构信息网络(HIN)已被提取用于提高推荐系统的性能,然而现有的提取方法没有考虑到向量自身的不同维度交互暗含的有用信息。为此,文章提出一种基于自外积的异构信息网络模型(HSopRec)用于改进推荐系统的性能。该模型能够通过自外积的方式有效地提取用户和物品原本暗含在异构信息网络的潜在关系。在世界开放商业数据集Yelp上进行的推荐性能的验证结果表明,与现有其他异构网络模型相比,HSopRec模型展现了更好的效果。

Network embedding has gained a lot of attention in BigData research in recent years,and heterogeneous information network(HIN)have been applied to improve the performance of recommendation system with various mining methods.However,the existing extraction methods do not take into account the useful information implied by the interaction of different dimensions of these vectors themselves,so we propose an improved recommendation model with self-outer product enhanced HIN(HSopRec),with which the potential relationship originally implied in the HIN between users and items can be effectively extracted through self-outer product.Compared with other similar models based on HINs,the HSopRec model shows very good results on the world’s open business data set Yelp.