Abstract:With the widespread use of online social media platforms, information diffusion has become a prevalent phenomenon, making Information Diffusion Prediction (IDP) increasingly important for various applications. Despite significant advancements in IDP research, existing methods often overlook issues of noise and sparsity in information diffusion data. User behaviors are frequently influenced by external factors, introducing noise into the data and hindering models' understanding of true diffusion patterns. Additionally, many users have limited interaction data, leading to data sparsity and restricting models' ability to effectively capture user preferences. To address these challenges, we propose a novel framework called DDiff, which tackles noise and sparsity issues through denoising diffusion and cross-domain contrastive learning. First, we introduce a graph learning encoder module that captures the social homophily of users through their relationships and higher-order connections via information diffusion hypergraphs (IDH). Next, a cross-domain contrastive learning module is designed to facilitate effective knowledge transfer between the information and social domains, addressing the sparsity problem. Furthermore, we propose a denoising diffusion module with IDH to effectively mitigate noise issues by introducing random noise in the forward process and iteratively recovering the corrupted embeddings in the reverse process. Finally, we implement a prediction module to determine the likelihood of subsequent users becoming infected. Experimental results demonstrate that DDiff significantly outperforms state-of-the-art methods in the information diffusion prediction task.

Information Diffusion Prediction with Latent Factor Disentanglement

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

Full-Scale Information Diffusion Prediction With Reinforced Recurrent Networks

H-Diffu: Hyperbolic Representations for Information Diffusion Prediction

Understanding Information Diffusion Via Heterogeneous Information Network Embeddings

A Diffusion Simulation User Behavior Perception Attention Network for Information Diffusion Prediction.

Improving Information Diffusion Prediction by Tackling Noise and Sparsity Challenges

A Novel Embedding Method for Information Diffusion Prediction in Social Network Big Data.

Multi-scale Information Diffusion Prediction with Reinforced Recurrent Networks

Modelling the Latent Semantics of Diffusion Sources in Information Cascade Prediction

Neural Information Diffusion Prediction with Topic-Aware Attention Network

MSIDP: Multi-scale Information Diffusion Prediction with Timestamp Information and Wide Dispersion

Independent Asymmetric Embedding for Information Diffusion Prediction on Social Networks

DyDiff-VAE: A Dynamic Variational Framework for Information Diffusion Prediction

Predicting Information Diffusion Using the Inter- and Intra-Path of Influence Transitivity.

Information Diffusion Prediction Via Cascade-Retrieved In-context Learning

Information Diffusion Prediction Based on Deep Attention in Heterogeneous Networks

MCDAN: a Multi-scale Context-enhanced Dynamic Attention Network for Diffusion Prediction

DyHGCN: A Dynamic Heterogeneous Graph Convolutional Network to Learn Users' Dynamic Preferences for Information Diffusion Prediction

Enhancing Multi-Scale Diffusion Prediction Via Sequential Hypergraphs and Adversarial Learning

A Dynamic Heterogeneous Graph Perception Network with Time-Based Mini-Batch for Information Diffusion Prediction