PRECIPITATION FORECASTING USING NEURAL NETWORK MODEL APPROACH
Atiek Iriany, Diana Rosyida, Agus Dwi Sulistyono, Budi Nurani Ruchjana
Department of Statistics, Faculty of Mathematics and Natural Science, University of Brawijaya, Malang, Indonesia
Department of Mathematics, Faculty of Mathematics and Natural Science, University of Padjajaran, Bandung, Indonesia
Abstract
Neural network constitutes a non-linear model requiring no statistical assumption. Along the development of which, neural network model has been frequently combined with time series and spatio temporal models. This current research combined neural network and spatio temporal models. One of spatio temporal models is GSTAR-SUR model. The weight projected in this current research is cross covariance normalized weight. This sort of weight is deemed suitable for data with high variability. The significant variable in GSTAR-SUR model containing cross covariance normalized weight was used as input layer of neural network model. The hidden layer made use of 10 neurons fulfilling the criteria of the lowest RMSE value and there was 1 neuron used as output. The data were in the form of 10-day precipitations in Junggo, Pujon, Tinjumoyo, and Ngujung, during the period of 2005 to 2014. This research has found out that NN-GSTAR-SUR model yielded better and more accurate forecasting, showing R^2 value of 61.77%.
Keywords: GSTAR, SUR, Neural Network, Cross Covariance Normalization
Topic: International Conference on Sustainable Agriculture