Meta-Graph Based Recommendation Fusion over Heterogeneous Information Networks
Huan Zhao (HKUST);Quanming Yao (HKUST);Jianda Li (HKUST);Yangqiu Song (HKUST);Dik Lee (HKUST)
Abstract
Heterogeneous Information Network (HIN) is a natural and general representation of data shown in modern large commercial recommender systems with heterogeneous types of data. Recommendation based on HIN faces two challenges: how to represent the high-level semantics of recommendation and how to fuse the heterogeneous information to make recommendation. In this paper, we solve the two challenges by first introducing the concept of meta-graph to HIN-based recommendation, and then solving the information fusion problem with a
