Comparing the Performance of Time Series Models for Forecasting Inflation Rate in the Sudan

Abstract

This paper empirically discussed the performance ability of five time series methods namely; trend equation, decomposition, exponential smoothing method and moving average, autoregressive models in modeling and forecasting inflation rates in the Sudan. Monthly readings of inflation rates data cover the period from January 2008 to July 2016 supplied by the Central Bureau of Statistics were used in the analysis of this paper. The five time series methods under consideration were applied to inflation rates data. MAPE, MAE and MSE accuracy measurements were being used in evaluation of the under consideration. The quadratic trend was chosen as an appropriate to fit inflation rate data using trend equation, additive decomposition method was selected as adequate model to represents decomposition method and double exponential smoothing model was chosen as a best model to represents exponential smoothing methods. Furthermore, the four time series models were evaluated based on accuracy measurements, the empirical findings conclude that an autoregressive model shown a smallest values of model selection criteria consequently it could be selected as an appropriate method for modeling and forecasting inflation rates in the Sudan, therefore for this particular data an autoregressive model is highly recommended.