A1215
Title: Online Bayesian model averaging for streaming data
Authors: Joyee Ghosh - The University of Iowa (United States) [presenting]
Aixin Tan - University of Iowa (United States)
Lan Luo - University of Iowa (United States)
Abstract: There is an increasing prevalence of streaming data generation in diverse fields like healthcare, finance, social media, and weather forecasting. In order to acquire helpful insights from these massive datasets, timely analysis is essential. The streaming data is assumed to be analyzed in batches. Traditional offline methods, which involve storing and analyzing all individual records, can be repeatedly applied to the cumulative data but encounter significant challenges in storage and computing costs. Existing online methods offer faster approximations, but most methods neglect model uncertainty, causing overconfidence and instability. To bridge this gap, novel online Bayesian approaches are proposed that incorporate model uncertainty within a Bayesian model averaging (BMA) framework, for generalized linear models (GLMs). Computationally efficient methods are proposed to update the posterior, with individual records from the latest batch of data and summary statistics from previous batches. Simulation studies and real data demonstrate that the methods can offer much faster analysis compared to traditional methods, with no substantial drop in accuracy.
