Time series forecasting with UCM model : a comparative study using the Tigris River data

Abstract EN

In this paper, we build two basic models to forecast a flow water of the Tigris river which enters to mosul city.

The first model is Unobserved Components Model which is abbreviated as UCM, the second is Autoregressive and Moving Average model which is mentioned as ARMA, we built 10 primary models from ARMA to data of the time series of flow Tigris river after we transfer the data to a standrized form to remove seasonal effects.

The best model which represented the data among ARMA models which are mentionted above is ARMA (2, 2) by depending on the correction of Akaike information criterion which is symbolized by AICc.

while ARMA(1, 2) model is the best model for forecasting because it has a minimum mean absolute error which is symbolized by MAE.

We obtained that the forecasting of flow water by UCM model is better than the results of ARMA (1, 2) model by depending on the criterion MAE.