Fourier-autoregressive (F-AR) coefficient non-linear time-series model for forecasting asymmetric cyclical data

Abstract

The aim of this study was to apply Fourier autoregressive (F-AR) model to describe and forecast asymmetric cyclical data. For carrying out statistical analysis, computer programs were developed using SAS, Ver. 9.2 software package. Twentysix years (1985–2010) quarterly oil sardine fish landings data (in tonnes) recorded at Central Marine Fisheries Research Institute, Kochi, India were used. Superiority of F- AR model over AR model was demonstrated by developing one-step ahead forecasts for two years’ hold-out data. Its potential use is to develop optimal import and export policies for Oil sardines. This type of information would also go a long way in enabling the Fishing industry in optimization of its resources. Efficient Oil sardine management strategies need to be evolved in order to allocate optimum number of boats and trawlers, etc.