The purpose of this study was to develop a framework that uses different forecasting methods and selects the best one with the least possible forecasting errors to predict harvesting of Largehead Hairtail (Trichiurus lepturus) stocks in the Persian Gulf and Sea of Oman. In this study, eleven different forecasting techniques including decomposition method (Multiplicative and Additive), moving average, exponential smoothing (Single, Double), trend an analysis (Linear, Exponential, Quadratic, S-Curve), and Winters method (Multiplicative and Additive) were performed by statistical technique to predict harvesting of T. lepturus stocks in the Persian Gulf and Sea of Oman. The results of model Quadratic Trend Analysis (MAPE=2.77, MAD=0.10, MSD=0.01) were better than other non-combined prediction models based on fewer error values and therefore a prediction was accomplished by using the same model for a period of five years. Various models used to identify orders of the autoregressive integrated moving average, ARIMA, (p, d, q) based on the AIC and BIC, and ARIMA (0, 1, 1) had the best fit with the process of changing T. lepturus annual landings in the southern coastal waters based on the selection criteria. According to the results,  the combined prediction model seems to be better than the non-combined prediction model, which predicts the future and its predictive numbers, indicating a low catch fluctuation of this species in Iranian southern waters.

Keywords: ARIMA, Catch, Largehead hairtail, prediction model

Full-Text [PDF 605 kb]   (796 Downloads)