Rainfall Intensity Forecast Using Ensemble Artificial Neural Network and Data Fusion for Tropical Climate

Abstract

This paper proposes an ensemble method based on neural network architecture and stacking generalization. This research objective is to develop a novel ensemble of Artificial Neural Network models with back propagation network and dynamic Recurrent Neural Network of nonlinear autoregressive network with exogenous inputs to improve prediction accuracy. Historical meteorological parameters and rainfall intensity have been used for predicting the rainfall intensity forecast. Hourly predicted rainfall intensity forecast are compared and analyzed for all models. The outcome shows that for 1 hour of prediction, the neural network ensemble forecast model returns 94 percent precision. The study concludes that the ensemble neural network model shows significant improvement in prediction performance as compared to the individual models.