Relational rule learning in decoupled heterogeneous subspaces

Service business now plays increasingly important role in real-world economy. This has stimulated the analytic requirement for generating insight from the structural and interrelated service data, so as to improve service operation and management excellence. In this paper, we propose a novel multi-relational classification algorithm, namely RSCC (Relational Subspace Collaborative Classification). RSCC restructures the relational dataset into a set of decoupled semantic-level subspaces while keeps the heterogeneity of relational data. It employs a heuristic rule learning strategy that globally searches for the best predicates effectively. Our experiments on multiple benchmark datasets demonstrate its performance and efficiency.