20 Sep 2013
VAR is multivariate time series method that can be used to analyzethe dynamic effects of disturbance factors contained in the system. VAR is amethod that used linear pattern in modeling. Sometimes if the data wassupposedly not linear, the linear method such as VAR will not suffice. Anotherapproach that suffice nonlinierity is neural network (NN). Feed forward neuralnetwork (NN) is the most commonly used archi-tecture in NN. In this paper,there are two methods conducted. They are VAR and the combination of VAR andNN, called VAR-NN. There are three variables involved in this study. They arerice production, rice procurement, and rice price. The comparison result byusing the rice procurement shows that VAR-NN yields better forecast than VAR.But, The comparison result by using the rice price data and rice procurementdata show that VAR-NN yields worse forecast than VAR. Since this study focusedon predicting rice procurement, so the best model achieved in this researchcomes from the class of VAR.
Copyrights © 2013