A distributed PSO-ARIMA-SVR hybrid system for time series forecasting

The combination of techniques in order to achieve more accurate predictions in time series forecasting has been widely applied. Statistical linar models such as the autoregressive integrated moving average (ARIMA) can not capture nonlinear patterns in time series. Therefore nonlinear models such as the support vector regression (SVR) are able to map such patterns. Thus time series can be decomposed in linear and nonlinear patterns. In order to capture both types of patterns a hybrid system comprised by ARIMA and SVR models optimized by the particle swarm optimization (PSO) algorithm is applied to perform predictions. The results show that the proposed method achieved promising results for one-step ahead predictions.