Commodity price prediction using neural network case study: Crude palm oil price

Commodities are the important factors in Indonesian economy. Being one of the biggest crude palm oil producers, the importance of knowing future crude palm oil price will bring significant impact to Indonesian economy. This paper describes an attempt to predict daily commodity prices, especially crude palm oil by employing neural network. First of all, this paper will explain how to construct dataset for learning and testing and how to build neural network model. There are several experiments to find what configuration of neural network should be selected to make the prediction more accurate, including testing two kinds of network topology named joint network and separated network. Proposed neural network model using joint network topology and regular normalization in momentum of 0.75 and learning rate of 0.05 is proven to be best model with minimum of 50000 iterations. Our proposed model has MAPE of 2.10 percent and RMSPE of 2.61 percent when tested using given experimental schemas.