Deep Popularity Prediction in Multi-Source Cascade with HERI-GCN

As a fundamental problem in social networks, popularity prediction is dedicated to predicting the scale of information diffusion, i.e., the number of involved users. Re- cently, deep learning methods for popularity prediction advance traditional approaches that rely on hand-crafted features. These deep learning methods employ automatic feature extraction to learn sequential and structural features based on either recurrent neural networks (RNNs) or graph neural networks (GNNs). How- ever, existing approaches ignore the multi-source cascade that consists of multiple sub-cascades originated from various source nodes with different content but under the same topic. Compared with single-source cascade, more cascading information can be observed from multi-source cascade and they are potentially correlated. How to correlate the diverse information and take advantage of them from both temporal and spatial aspects is critical for prediction. To this end, we propose a novel framework, called HEt- erogeneous Recurrent Integrated Graph Convolutional Neural Network (HERI-GCN). Specifically, we first construct a het- erogeneous cascade graph to model the multi-source cascade where time intervals are treated as heterogeneous time nodes to connect multiple sub-cascades as a whole. After that, we propose a heterogeneous GCN to learn rich features from multi-source cascade. RNN is organically integrated into the heterogeneous GCN to overcome the limited learning ability toward temporal and spatial data. We evaluate HERI-GCN through comparative experiments on three datasets from Sina-Weibo and Twitter platforms. The experimental evaluation shows that HERI-GCN outperforms the state-of-the-art baseline methods.