Popularity Prediction on Social Platforms with Coupled Graph Neural Networks (1906.09032v2)

Abstract: Predicting the popularity of online content on social platforms is an important task for both researchers and practitioners. Previous methods mainly leverage demographics, temporal and structural patterns of early adopters for popularity prediction. However, most existing methods are less effective to precisely capture the cascading effect in information diffusion, in which early adopters try to activate potential users along the underlying network. In this paper, we consider the problem of network-aware popularity prediction, leveraging both early adopters and social networks for popularity prediction. We propose to capture the cascading effect explicitly, modeling the activation state of a target user given the activation state and influence of his/her neighbors. To achieve this goal, we propose a novel method, namely CoupledGNN, which uses two coupled graph neural networks to capture the interplay between node activation states and the spread of influence. By stacking graph neural network layers, our proposed method naturally captures the cascading effect along the network in a successive manner. Experiments conducted on both synthetic and real-world Sina Weibo datasets demonstrate that our method significantly outperforms the state-of-the-art methods for popularity prediction.

• The paper proposes a novel CoupledGNN framework that models user activation and influence propagation to predict online content popularity.
• It leverages state and influence gating functions to capture dynamic cascading effects, yielding over 10% improvement on Sina Weibo data.
• The method demonstrates adaptability with iterative neural layers and diverse datasets, offering new insights for network-based diffusion analysis.

Popularity Prediction on Social Platforms with Coupled Graph Neural Networks

The paper "Popularity Prediction on Social Platforms with Coupled Graph Neural Networks" addresses the task of predicting the future popularity of online content by explicitly modeling the cascading effect in information diffusion. This work introduces a novel approach utilizing coupled graph neural networks (GNNs) to represent the interplay between user activation and influence spread within a social network.

Introduction

Many traditional methods for predicting the popularity of online content rely on heuristics related to demographics, temporal patterns, and structural features of early adopters. However, they tend to inadequately account for the cascading influence, which is critical for accurate prediction. The paper proposes a network-aware approach, recognizing the dynamic interaction between activated nodes (users who have adopted the information) and their potential to influence neighboring nodes.

Coupled Graph Neural Networks (CoupledGNN)

The proposed solution involves CoupledGNN, which uses two graph neural networks to separately model activation states and influence propagation:

1. State Graph Neural Network: Models the activation propagation, effectively simulating how an active user may trigger their inactive neighbors through the network.
2. Influence Graph Neural Network: Captures the spread of interpersonal influence, which is gated by users' activation states.

These networks are coupled through gating mechanisms that allow the influence from neighbors to depend on both the intrinsic state and network-driven influence, capturing complex diffusion dynamics.

Methodology

The CoupledGNN framework is developed with iterative layers, where each layer models the influence and activation interplay over the network. The influence gating function allows adaptive weighting of neighboring influences, and the state gating functions provide non-linear filtering of this influence.

The model assesses network-based representations and can integrate additional data types like temporal features when available. This versatility enhances its accuracy across different datasets and settings.

Experimental Results

Experiments conducted on synthetic and Sina Weibo datasets demonstrate the model's superiority in prediction accuracy over existing state-of-the-art methods. The CoupledGNN approach significantly outperforms baselines such as feature-based methods, DeepCas, and SEISMIC, showing more than 10% improvement in prediction metrics on the Sina Weibo dataset.

Implications and Future Work

The implications of this work extend beyond popularity prediction; the architectural innovation of using coupled GNNs can be leveraged in other domains requiring dynamic influence and state modeling across networked nodes. Future developments may focus on incorporating more complex temporal and contextual information into the network structure, as well as extending the methodology to accommodate real-time predictions on evolving networks.

Conclusion

The incorporation of coupled graph neural networks opens a promising direction for accurately predicting content popularity by encompassing the complex dynamics of information diffusion on social platforms. By effectively modeling the cascading effect, this approach lays groundwork for applications requiring nuanced predictions of network influence. The CoupledGNN stands as a compelling advancement in the field of social network analysis and predictive modeling.