Uncovering and predicting the dynamic process of information cascades with survival model
详细信息 查看全文
- 作者：Linyun Yu ; Peng Cui ; Fei Wang ; Chaoming Song…
- 关键词：Information cascades ; Social network ; Dynamic processes prediction
- 刊名：Knowledge and Information Systems
- 出版年：2017
- 出版时间：February 2017
- 年：2017
- 卷：50
- 期：2
- 页码：633-659
- 全文大小：
- 刊物类别：Computer Science
- 刊物主题：Information Systems and Communication Service; IT in Business;
- 出版者：Springer London
- ISSN：0219-3116
- 卷排序：50
文摘
Cascades are ubiquitous in various network environments. Predicting these cascades is decidedly nontrivial in various important applications, such as viral marketing, epidemic prevention, and traffic management. Most previous works have focused on predicting the final cascade sizes. As cascades are dynamic processes, it is always interesting and important to predict the cascade size at any given time, or to predict the time when a cascade will reach a certain size (e.g., the threshold for an outbreak). In this paper, we unify all these tasks into a fundamental problem: cascading process prediction. That is, given the early stage of a cascade, can we predict its cumulative cascade size at any later time? For such a challenging problem, an understanding of the micromechanism that drives and generates the macrophenomena (i.e., the cascading process) is essential. Here, we introduce behavioral dynamics as the micromechanism to describe the dynamic process of an infected node’s neighbors getting infected by a cascade (i.e., one-hop sub-cascades). Through data-driven analysis, we find out the common principles and patterns lying in the behavioral dynamics and propose the novel NEtworked WEibull Regression model for modeling it. We also propose a novel method for predicting cascading processes by effectively aggregating behavioral dynamics and present a scalable solution to approximate the cascading process with a theoretical guarantee. We evaluate the proposed method extensively on a large-scale social network dataset. The results demonstrate that the proposed method can significantly outperform other state-of-the-art methods in multiple tasks including cascade size prediction, outbreak time prediction, and cascading process prediction.