Graph-Enhanced Multi-Task Learning of Multi-Level Transition Dynamics for Session-based Recommendation

Session-based recommendation plays a central role in wide spectrum of online applications, ranging from e-commerce to advertising services. However, the majority existing session-based techniques (e.g., attention-based recurrent network or graph neural network) are not well-designed for capturing complex transition dynamics exhibited with temporally-ordered and multi-level interdependent relation structures. These methods largely overlook hierarchy item transitional patterns. In this paper, we propose multi-task learning framework Multi-level Transition Dynamics (MTD), which enables jointly intra- inter-session automatic hierarchical manner. Towards end, first develop position-aware attention mechanism learn regularities within individual session. Then, graph-structured encoder is proposed explicitly capture cross-session transitions form high-order connectivities by performing embedding propagation global context. The process integrated, preserve underlying low- high-level relationships common latent space. Extensive experiments on three real-world datasets demonstrate superiority MTD as compared state-of-the-art baselines.