Recommendation in heterogeneous information networks with implicit user feedback

Abstract

Recent studies suggest that by using additional user or item relationship information when building hybrid recommender systems, the recommendation quality can be largely improved. However, most such studies only consider a single type of relationship, e.g., social network. Notice that in many applications, the recommendation problem exists in an attribute-rich heterogeneous information network environment. In this paper, we study the entity recommendation problem in heterogeneous information networks. We propose to combine various relationship information from the network with user feedback to provide high quality recommendation results.The major challenge of building recommender systems in heterogeneous information networks is to systematically define features to represent the different types of relationships between entities, and learn the importance of each relationship type. In the proposed framework, we first use meta-path-based latent features to represent the connectivity between users and items along different paths in the related information network. We then define a recommendation model with such latent features and use Bayesian ranking optimization techniques to estimate the model. Empirical studies show that our approach outperforms several widely employed implicit feedback entity recommendation techniques.

A survey of heterogeneous information network analysis

“…3(c), a bibliographic information network is a typical star-schema heterogeneous network [14], [13], containing different objects (e.g., paper, venue, author, and term) and links among them. Many other datasets can also be represented as star-schema networks, such as the movie data [34], [35] from the Internet Movie Database 2 (IMDB) and the patent data [36] from US patents data…”

“…[ 35], [133] introduce meta-path-based latent features to represent the connectivity between users and items along different types of paths, and then define recommendation models at both global and personalized levels with Bayesian ranking optimization techniques. Also based on meta path, Burke et al [134] present an approach for recommendation which incorporates multiple relations in a weighted hybrid.…”

“…Utilizing different contexts information, Jamali et al [31] propose a contextdependent matrix factorization model which considers a general latent factor for every entity and context-dependent latent factors for every context. Using user implicit feedback data, Yu et al [35], [133] solve the global and personalized entity recommendation problem. Based on related interest groups, Ren et al [135] propose a cluster-based citation recommendation framework to predict each query's citations in bibliographic networks.…”

A Fuzzy-Based Personalized Recommender System for Local Businesses

“…[16] builds an attribute-rich heterogeneous information network of Yelp reviews and combines various related information from the network with user feedback to provide high-quality recommendations for users. While much prior research has focused on providing recommendations for users based on their similar interests in businesses, social structure, or participation in reviewing systems, little attention has been paid to the relationships between businesses.…”

“…The rich data from these systems provides plenty of opportunities to provide meaningful personalized recommendation results, based on users' historical data. Different approaches have arisen to study personalized recommender systems like these, such as those that are based on user interests [11], social factors [5] and heterogeneous networks [16] to help users browse, search, and explore the system, given their own preference of information. However, little attention has been paid to the recommendations from the business side.…”