APPLICATION OF AN ARTIFICIAL NEURAL NETWORK (ANN) MODEL FOR PREDICTING OIL PALM FRESH FRUIT BUNCH (FFB) YIELD BASED ON RAINFALL AND THE YIELD PREVIOUSLY

Abstract

Abstract. To predict oil palm yield in 2018 at 4 Indonesian Oil Palm Research Institue field Trial Plantation (Padang Mandarsah, Dalu-dalu, Bukit Sentang, and Aek Pancur), then it was built an Artificial Neuron Network (ANN) model. The data used were monthly yield and rainfall during 2013-2017. The model output taking by the relation of non-linear Autoregressive to  the rainfall external input (NARX). The model built processing including training using the data 2013-2015, validation using the data 2016, testing using the data 2017. From the testing model result, were taken a good fit model architecture n-d-h-o (variable input,n ; d-tapped delayed , d, node hidden, h; output layer, o) and correlation coefficient (r) between output model and actual data for each plantation. Padang Mandarsah 2-3-4-1 with r= 0,84 ; Dalu-dalu 2-24-5-1 with r = 0,74; Bukit Sentang 2-24-10-1 with r = 0,84, and Aek Pancur 2-3-5-1 with r = 0,86.