MDBF: Meta-Path-Based Depth and Breadth Feature Fusion for Recommendation in Heterogeneous Network

The main challenge of recommendation in a heterogeneous information network comes from the diversity nodes and links problem semantic expression ambiguity caused by diversity. Therefore, we propose movie algorithm for called Meta-Path-Based Depth Breadth Feature Fusion(MDBF). Using random walk depth feature learning, can extract meta-path that reflects overall structure network. In addition, using walks adjacent nodes, breadth meta-path, preserving neighborhood node. If there is some auxiliary information, it will be learned its own meta-paths. Then, all sequences fused Skip-gram to obtain final vector. process, based on traditional collaborative filtering, secondary filtering recommendation. experimental results show that, without external compared existing state-of-the-art models, improves each index an average 12% MovieLens 22% MovieTweetings. not only effect recommendation, but also provides application scenarios accurate services through information.