The goal of medication recommendation system is to recommend appropriate pharmaceutical interventions based on a patient’s diagnosis. Most of existing approaches often formulate these recommendations use data on diagnoses, procedures, and prescriptions accumulated in the electronic health records (EHR), and despite the great successes, they seem to have limitations on modelling the significance of medication to a patient’s current visit and mining fine-grained medication representation information. To address these issues, we propose a novel Significanceaware Medication Recommendation (SMRec) framework built on significance of medication to patients and fine-grained medication representation learning. Specifically, we first design a encoding mechanism to compute significance information of medications for each patient’s visit. Then, we utilize the set-level medication co-occurrence graph based on patients’ medical history which integrates temporal dependency to learn fine-grained medication representations. Experimental results on the publicly available MIMIC-III dataset demonstrate the superior effectiveness of our model compared to other approaches1.

Sequential recommendation has been developed to predict the next item in which users are most interested by capturing user behavior patterns embedded in their historical interaction sequences. However, most existing methods appear to exhibit limitations in modeling fine-grained dependencies embedded in users’ various periodic behavior patterns and heterogeneous dependencies across multi-behaviors. Towards this end, we propose a Filter-enhanced Hypergraph Transformer framework for Multi-Behavior Sequential Recommendation (FHT-MB) to address the above challenges. Specifically, a multi-scale filter layer equipped with multi-learnable filters is devised to encode behavior-aware sequential patterns emerging from different periodic trends (e.g., daily or weekly routines), and then a hypergraph structure is devised to extract heterogeneous dependencies across users’ multiple types of behaviors. Extensive experiments on two real-world e-commerce datasets show the superiority of our proposed FHT-MB over various state-of-the-art methods.1

Next basket recommender systems (NBRs) aim to recommend a user’s next (shopping) basket of items via modeling the user’s preferences towards items based on the user’s purchase history, usually a sequence of historical baskets. Due to its wide applicability in the real-world E-commerce industry, the studies NBR have attracted increasing attention in recent years. NBRs have been widely studied and much progress has been achieved in this area with a variety of NBR approaches having been proposed. However, an important issue is that there is a lack of a systematic and unified evaluation over the various NBR approaches. Different studies often evaluate NBR approaches on different datasets, under different experimental settings, making it hard to fairly and effectively compare the performance of different NBR approaches. To bridge this gap, in this work, we conduct a systematical empirical study in NBR area. Specifically, we review the representative work in NBR and analyze their cons and pros. Then, we run the selected NBR algorithms on the same datasets, under the same experimental setting and evaluate their performances using the same measurements. This provides a unified framework to fairly compare different NBR approaches. We hope this study can provide a valuable reference for the future research in this vibrant area.

Session-based recommendations (SBR) play an important role in many real-world applications, such as e-commerce and media streaming. To perform accurate session-based recommendations, it is crucial to capture both sequential dependencies over a sequence of adjacent items and complex item transitions over a set of items within sessions. Note that item transitions are not necessarily dependent on sequential dependencies, e.g., the transition from one item to the other distant item in a session is often not sequential. However, almost all the existing session-based recommender systems (SBRS) fail to consider both kinds of information, which leads to their limited performance improvement. Aiming at this deficiency, we propose a novel sequential dependency enhanced graph neural network (SDE-GNN) to capture both sequential dependencies and item transition relations over items within sessions for more accurate next-item recommendations. Specifically, we first devise a sequential dependency learning module to capture the sequential dependencies over a sequence of adjacent items in each session. Then, we propose an item transition learning module to capture complex transitions between items. In the module, a novel residual gate and a specialized attention mechanism are integrated into gate-GNN to build an attention augmented GNN, called AU-GNN. Finally, we devise a gated fusion component to combine the learned sequential dependencies and item transitions together in preparation for the subsequent next-item recommendations. Exhaustive experiments on two public real-world data sets demonstrate the superiority of SDE-GNN over the state-of-the-art methods.