Entity Recommendation Via Integrating Multiple Types of Implicit Feedback in Heterogeneous Information Network

Recently, heterogeneous information network(HIN) analysis has attracted a lot of attentions. One of the HIN application is recommendation. Due to HIN containing multiple different objects and links and rich semantic meanings, it is promising to generate better recommendation. Previous studies on movie recommendation have combined the single implicit feedback information with heterogeneous information network to create an efficient recommendation. In this paper, we combined multiple types of implicit feedback data with heterogeneous information network to achieve better movie recommendation. We propose the latent features of multiple types implicit feedback matrix along different types of meta path to connect users and movies. We define a recommendation model and use Bayesian ranking optimization techniques to estimate the proposed model. Empirical studies on Douban dataset show that our approach can make better recommendation than previous works.