Abstract:
In this work two time series, one representing Sudan annual fish production and the other the monthly fish supply in Khartoum markets, were studied. The classical decomposition method was first used, after checking for stationarity of both series, to detect and to estimate any trend or seasonal component that may be present in the fish supply series. Spectral analysis was also used to shed more light on the presence of cyclical patterns in the two series. On the time domain, an attempt was made to build mathematical models for fish production and fish monthly supply that may be used to forecast future production and supply. For this end, several models, including ARIMA and Exponential Smoothing were tried and the best among these was determined. Application of the decomposition method to the fish supply revealed an apparent seasonal effect in certain months. Spectral analysis showed that the dominance of a single cycle of length of about 7 years in the fish production series and a cycle of length 12 months in the fish supply series confirmed the decomposition method results. On the other hand, examination of the different models showed that the best forecasting model for fish product and supply series were the Quadratic Trend Model and Winter’s Exponential Smoothing model, respectively. Forecasts for both series were carried out.
