Learning User-Specific Latent Influence and Susceptibility from Information Cascades

• DOI: 10.1609/aaai.v29i1.9213
• Publication Date: 2022-06-23T23:16:13Z
• Authors: Yongqing Wang; Huawei Shen; Shenghua Liu; Xueqi Cheng
• Keywords: Overfitting; Information cascade; Microblogging; Viral marketing; Limiting

Predicting cascade dynamics has important implications for understanding information propagation and launching viral marketing. Previous works mainly adopt a pair-wise manner, modeling the probability between pairs of users using n2 independent parameters n users. Consequently, these models suffer from severe overfitting problem, especially without direct interactions, limiting their prediction accuracy. Here we propose to model by learning two low-dimensional user-specific vectors observed cascades, capturing influence susceptibility respectively. This requires much less thus could combat problem. Moreover, this naturally context-dependent factors like cumulative effect in propagation. Extensive experiments on synthetic dataset large-scale microblogging demonstrate that outperforms existing at predicting dynamics, size, "who will be retweeted."