Fitting Nonlinear Time-series Model Using Swarm Optimization Technique

The well-known Box-Jenkins’ Autoregressive Integrated Moving Average (ARIMA) methodology for fitting time-series data has some major limitations. To this end, Exponential Autoregressive (EXPAR) family of models may be employed. An important characteristic feature of EXPAR is that it is capable of modelling those data sets that depict cyclical variations. Further, it can also be used when data show non-Gaussianity. In this paper, methodology for fitting EXPAR through powerful optimization tool called Particle Swarm optimization (PSO) is described in detail. As an illustration, PSO is used for fitting EXPAR model in India’s annual lac export data. Moreover, the performance of fitted EXPAR model is compared with ARIMA from modelling and forecasting point of view. It is concluded that the performance of EXPAR model is better than ARIMA for the dataset under consideration.