Compared to the predicted value of the residual Grey and ANN method, we can see that the prediction accuracy is improved. This shows the limitations of the residual Grey method that added an almost linear error correction to the predicted value. By estimating error values occurring in process of time series using ANN, a nonlinear error prediction value caused prediction accuracy to increase. In previous studies, prediction by ANN cannot describe in the blackbox, have had difficulty in interpretation. On the other hand, as ANN method was applied to the only error term in this hybrid method, we can reduce the unexplained part because of blackbox and increase the prediction accuracy.
展开▼
