Abstract:Human mobility prediction is a fundamental task essential for various applications, including urban planning, location-based services and intelligent transportation systems. Existing methods often ignore activity information crucial for reasoning human preferences and routines, or adopt a simplified representation of the dependencies between time, activities and locations. To address these issues, we present Hierarchical Graph Attention Recurrent Network (HGARN) for human mobility prediction. Specifically, we construct a hierarchical graph based on all users' history mobility records and employ a Hierarchical Graph Attention Module to capture complex time-activity-location dependencies. This way, HGARN can learn representations with rich human travel semantics to model user preferences at the global level. We also propose a model-agnostic history-enhanced confidence (MAHEC) label to focus our model on each user's individual-level preferences. Finally, we introduce a Temporal Module, which employs recurrent structures to jointly predict users' next activities (as an auxiliary task) and their associated locations. By leveraging the predicted future user activity features through a hierarchical and residual design, the accuracy of the location predictions can be further enhanced. For model evaluation, we test the performances of our HGARN against existing SOTAs in both the recurring and explorative settings. The recurring setting focuses on assessing models' capabilities to capture users' individual-level preferences, while the results in the explorative setting tend to reflect the power of different models to learn users' global-level preferences. Overall, our model outperforms other baselines significantly in all settings based on two real-world human mobility data benchmarks. Source codes of HGARN are available at <a class="link-external link-https" href="https://github.com/YihongT/HGARN" rel="external noopener nofollow">this https URL</a>.

DisenHCN: Disentangled Hypergraph Convolutional Networks for Spatiotemporal Activity Prediction

Tackling higher-order relations and heterogeneity: Dynamic heterogeneous hypergraph network for spatiotemporal activity prediction

DisenHAN: Disentangled Heterogeneous Graph Attention Network for Recommendation

HSDN: A High-Order Structural Semantic Disentangled Neural Network

A Hypergraph-based Hybrid Graph Convolutional Network for Intracity Human Activity Intensity Prediction and Geographic Relationship Interpretation

Dynamic Hypergraph Convolutional Network

Disentangled Hypergraph Collaborative Filtering for Social Recommendation

GSEN: an Ensemble Deep Learning Benchmark Model for Urban Hotspots Spatiotemporal Prediction.

Disentangled Hyperbolic Representation Learning for Heterogeneous Graphs

Discrete-time Temporal Network Embedding via Implicit Hierarchical Learning in Hyperbolic Space

Activity-aware Human Mobility Prediction with Hierarchical Graph Attention Recurrent Network

Disentangled Graph Convolutional Networks

STHGCN: A spatiotemporal prediction framework based on higher-order graph convolution networks

Disentangled Graph Social Recommendation

LGD-GCN: Local and Global Disentangled Graph Convolutional Networks

Predicting User Activity Intensity Using Geographic Interactions Based on Social Media Check-In Data

HDM-GNN: A Heterogeneous Dynamic Multi-view Graph Neural Network for Crime Prediction

Spatial-Temporal Sequential Hypergraph Network for Crime Prediction with Dynamic Multiplex Relation Learning

Enhancing Information Diffusion Prediction with Self-Supervised Disentangled User and Cascade Representations

Predicting Activity and Location with Multi-task Context Aware Recurrent Neural Network.

Disentangled Modeling of Social Homophily and Influence for Social Recommendation