Integrated Spatio-temporal Forecasting Method by DRNN and ARIMA Combined Model
Zhixiong Mei · Journal of Chinese Computer Systems · 2010
After reviewing the current spatio-temporal forecasting researches,this paper proposes a spatio-temporal integration forecasting approach based on dynamic recurrent neural network(DRNN) and autoregressive integrated moving average(ARIMA) combined model.The approach first forecasts time series by ARIMA model,then captures the hidden spatial correlation by DRNN between spatio-temporal data,and finally combines the individual temporal and spatial forecasting results based on linear regression to produce the final spatio-temporal integration forecasting result.Experimental results show that the approach can obtain better performance and forecasting precision than the approaches that don't consider spatial correlation or monomial forecasting method,can forecast dynamic process over space effectively in virtue of its strongly dynamic handling and computing abilities.