A Semiparametric Regression Ensemble Model for Rainfall Forecasting Based on RBF Neural Network

Abstract

Rainfall forecasting is very important research topic in disaster prevention and reduction. In this study, a semiparametric regression ensemble (SRE) model is proposed for rainfall forecasting based on radial basis function (RBF) neural network. In the process of ensemble modeling, original data set are partitioned into some different training subsets via Bagging technology. Then a great number of single RBF neural network models generate diverse individual neural network ensemble by training subsets. Thirdly, the partial least square regression (PLS) is used to choose the appropriate ensemble members. Finally, SRE is used for neural network ensemble for prediction purpose. Empirical results obtained reveal that the prediction using the SRE model is generally better than those obtained using the other models presented in this study in terms of the same evaluation measurements. Our findings reveal that the SRE model proposed here can be used as a promising alternative forecasting tool for rainfall to achieve greater forecasting accuracy and improve prediction quality further.