Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

HierCas: Hierarchical Temporal Graph Attention Networks for Popularity Prediction in Information Cascades

Zhizhen Zhang, Xiaohui Xie, Yishuo Zhang, Lanshan Zhang, Yong Jiang

TL;DR

HierCas introduces a non-sampling, dynamic graph framework for macro-level cascade popularity prediction. It fuses time-aware node embeddings, temporal graph attention, and hierarchical pooling to model the entire cascade graph and predict the increment ΔP after observation time. The approach yields consistent gains over strong baselines on two real-world datasets, with ablations confirming the importance of time/size embeddings and multi-level pooling. This work advances cascade modeling by preserving continuous diffusion dynamics and jointly leveraging temporal and structural information, enabling more accurate and interpretable popularity forecasts in information cascades.

Abstract

Information cascade popularity prediction is critical for many applications, including but not limited to identifying fake news and accurate recommendations. Traditional feature-based methods heavily rely on handcrafted features, which are domain-specific and lack generalizability to new domains. To address this problem, researchers have turned to neural network-based approaches. However, most existing methods follow a sampling-based modeling approach, potentially losing continuous dynamic information that emerges during the information diffusion process. In this paper, we propose Hierarchical Temporal Graph Attention Networks for cascade popularity prediction (HierCas), which operates on the entire cascade graph by a dynamic graph modeling approach. By leveraging time-aware node embedding, graph attention mechanisms, and hierarchical pooling structures, HierCas effectively captures the popularity trend implicit in the complex cascade. Extensive experiments conducted on two real-world datasets in different scenarios demonstrate that our HierCas significantly outperforms the state-of-the-art approaches. We have released our code at https://github.com/Daisy-zzz/HierCas.

HierCas: Hierarchical Temporal Graph Attention Networks for Popularity Prediction in Information Cascades

TL;DR

HierCas introduces a non-sampling, dynamic graph framework for macro-level cascade popularity prediction. It fuses time-aware node embeddings, temporal graph attention, and hierarchical pooling to model the entire cascade graph and predict the increment ΔP after observation time. The approach yields consistent gains over strong baselines on two real-world datasets, with ablations confirming the importance of time/size embeddings and multi-level pooling. This work advances cascade modeling by preserving continuous diffusion dynamics and jointly leveraging temporal and structural information, enabling more accurate and interpretable popularity forecasts in information cascades.

Abstract

Information cascade popularity prediction is critical for many applications, including but not limited to identifying fake news and accurate recommendations. Traditional feature-based methods heavily rely on handcrafted features, which are domain-specific and lack generalizability to new domains. To address this problem, researchers have turned to neural network-based approaches. However, most existing methods follow a sampling-based modeling approach, potentially losing continuous dynamic information that emerges during the information diffusion process. In this paper, we propose Hierarchical Temporal Graph Attention Networks for cascade popularity prediction (HierCas), which operates on the entire cascade graph by a dynamic graph modeling approach. By leveraging time-aware node embedding, graph attention mechanisms, and hierarchical pooling structures, HierCas effectively captures the popularity trend implicit in the complex cascade. Extensive experiments conducted on two real-world datasets in different scenarios demonstrate that our HierCas significantly outperforms the state-of-the-art approaches. We have released our code at https://github.com/Daisy-zzz/HierCas.
Paper Structure (20 sections, 13 equations, 4 figures, 2 tables)

This paper contains 20 sections, 13 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Preliminaries
  4. Methodology
  5. Temporal Graph Attention
  6. Time-aware Node Embedding
  7. Graph Attention Mechanism
  8. Multi-level Graph Pooling
  9. Experiments
  10. Datasets
  11. Experimental Setups
  12. Evaluation metrics
  13. Baselines and Parameter Settings
  14. Overall Performance
  15. Ablation Study
  16. ...and 5 more sections

Figures (4)

  • Figure 1: Detailed framework of HierCas. (a) is the main architecture. HierCas inputs the observed cascade graph, stacking several temporal graph attention layers to model temporal and structural information. Then the multi-level graph pooling module aggregates different localities learned by graph attention layers to make graph-level predictions. (b) takes $v_6$ as an example to show the node aggregation process in each layer, including the time-aware node embedding (time embedding RGB]255,242,204$\Phi(\cdot)$ and size embedding RGB]226,240,217$S(\cdot)$) and graph attention mechanisms.
  • Figure 2: Impact of the number of graph attention layers.
  • Figure 3: Impact of the number of neighbors.
  • Figure 4: Visualization of nodes' contributions.

Theorems & Definitions (2)

  • Definition 1: Cascade Graph
  • Definition 2: Cascade Popularity Prediction