Determination of the most appropriate statistical method for estimating the production values of medium density fiberboard

Abstract

This study determines an optimum method to predict Turkish Medium Density Fiberboard (MDF) production values using ARIMA (Box-Jenkins), regression, and Artificial Neural Network (ANN). The prediction performance of these methods is also compared. A total of 14 independent variables, likely to influence MDF production, were determined, and the production values of the next 9 years (2017-2025) were predicted on the basis of these variables. The test results indicate that the best Mean Squared Error (MSE), Mean Absolute Percentage Error (MAPE), and Mean Absolute Deviation (MAD) prediction performance belongs to the prediction performed with ANN.