NS Seminar: Paper Presentation

Date and Location

Feb 29, 2016 - 2:00pm to 3:00pm
Network Science Lab, Bldg 434, Rm 122

Speaker

Hannah Wolfe
Computer Science Trainee

Abstract

From Micro to Macro: Uncovering and Predicting Information Cascading Process with Behavioral Dynamics

Authors: Linyun Yu, Peng Cui, Fei Wang, Chaoming Song, Shiqiang Yang

Presented by: Hannah Wolfe

Summary:  A presentation on cascading process prediction, focused on predicting final cascade time, cascade peak time and cascade size at any moment.  The Networked Weibull Regression model is proposed for behavioral dynamic modeling of survival function.  A process is proposed to selectively aggregate behavioral dynamics and a scalable solution to approximate the cascading process.  The parameters of the survival function for hidden nodes can be predicted based off of it's neighbors using linear regression.  The solution outperforms other state of the art baselines for cascade size prediction, outbreak time prediction and cascade process prediction.  This process was extensively tested on a large Weibo social network dataset.