A0774
Title: Variational Bayesian inference for network autoregression models
Authors: Wei-Ting Lai - National Central University (Taiwan) [presenting]
Abstract: A variational Bayesian (VB) method is developed for estimating large-scale dynamic network models in a network autoregressive framework. The proposed VB method allows automatic identification of the dynamic structure of such a model and obtains a direct approximation of the posterior density. Compared to Markov Chain Monte Carlo (MCMC) based sampling methods, the VB method improves computational efficiency without losing estimation accuracy. In real-world data analysis, we apply the proposed VB algorithm to day-ahead natural gas flow forecasting for the German gas transmission network with 51 nodes from October 2013 to September 2015. The VB method provides promising prediction accuracy as well as detected structural dynamic dependencies.
