Nonlinear time series forecasting with Bayesian neural networks
作者：
Highlights：
• Evolutionary Monte Carlo algorithm is proposed to train Bayesian neural networks.
• The proposed approach is based on Gaussian approximation of Bayesian learning.
• Monte Carlo methods is integrated with GA and fuzzy membership functions.
• Time series forecasting was made over the weekly sales of a Finance Magazine.
• All the methods were compared in terms of forecasting performance on the test data.
摘要
•Evolutionary Monte Carlo algorithm is proposed to train Bayesian neural networks.•The proposed approach is based on Gaussian approximation of Bayesian learning.•Monte Carlo methods is integrated with GA and fuzzy membership functions.•Time series forecasting was made over the weekly sales of a Finance Magazine.•All the methods were compared in terms of forecasting performance on the test data.
论文关键词：Nonlinear time series,Bayesian neural networks,Gaussian approximation,Recursive hyperparameters,Genetic algorithms,Hybrid Monte Carlo simulations
论文评审过程：Available online 9 May 2014.
论文官网地址：https://doi.org/10.1016/j.eswa.2014.04.035
