Bayesian Neural Network Ensemble Model based on Partial Least Squares Regression and Its Application in Rainfall Forecasting

摘要

Rainfall forecasting is an essential tool in order to reduce the risk to life and alleviate economic losses. In this paper, using Bayesian techniques design a neural network ensemble model for rainfall forecasting. Firstly, using Bagging techniques and the different neural network algorithm are applied so as to generate an ensemble individual, and then the Partial Least Square regression technique are used to extract the ensemble members. Finally, Bayesian Neural Network is used to ensemble for rainfall forecasting model. The proposed approach is applied to real rainfall data. Our findings reveal that the Bayesian Neural Network ensemble model proposed here can greatly improve the modelling forecasting for a Meteorological application.