Comparison of Some Forecasting Time Series Models Using Cotton Productivity Data in the Sudan for the period 1953-2007

Abstract

This research aims to build statistical models for cotton productivity in Sudan (1953- 2007) using exponential smoothing model, classical moving average model and Box-Jenkins models of time series analysis to select the best model among them according to some statistical criteria (mean error square, standard error of mean, mean absolute error). The application for this study showed that the exponential smoothing model is the best model to forecast cotton yield in the Sudan according to the lowest value of mean square error and lowest standard error of mean. Also we used Kolmogorov-Smirnov test to test whether the noise residuals are normally distributed. The study recommended the following:- –To give importance to the forecasting study it can be useful in future planning to face change that may happen in the future. – Trying to apply other models may be better in forecasting. – To give importance to data quality because it is the sours of the modeling