Benchmarking Bayesian neural networks for time series forecasting

Rafael A. Calvo
SEDAL,University of Sydney
rafa@sedal.usyd.edu.au
Marwan Jabri
SEDAL, University of Sydney
marwan@sedal.usyd.edu.au

Abstract:

We report a benchmarking of neural networks and regression techniques in a time series forecasting task. The estimation errors, computing costs and additional information obtained by Bayesian neural networks are compared with other neural network models and with Multivariate Adaptive Regression Splines (MARS). The Mackey Glass time series in chaotic regime was used to generate the two data sets, one with noise and one without.

• Introduction
  • Bayesian Learning and forecasting
  • Markov Chain Monte Carlo
• Data and Methods
• Results
  • Estimation
  • Computational expenses
  • Prediction intervals
• Conclusions
• References
• About this document ...