Prediction of crude palm oil production in Indonesia using nonlinear autoregressive exogenous neural network with genetic algorithm

Abstract

Artificial neural network (ANN) has widely used as time series prediction tool. This study proposes one type of ANN, namely the nonlinear autoregressive exogenous neural network (NARX NN) model to predict crude palm oil production. The NARX NN model involves nonlinear activation functions both in hidden and output layers and a feedback link network from the output layer. The weights of the NARX NN model are optimized using Genetic Algorithms (GA). We employ the method on annual crude palm oil production in Indonesia from 1970 to 2019 on the training and testing data set. The NARX NN with GA optimization delivers more accurate predictions than without GA on training data. But both methods deliver almost the same performance on testing data.