Elman rolling or sliding window in time series forecasting
in [Matlab]
|
From: Jorge Pasquotto on 11 Apr 2010 20:55 Dear all, There are two approaches in updating forecasting models over time: the rolling and moving(sliding) window. The rolling window approach uses all available data to train neural networks while the moving window approach uses a set of most recent observations to estimate the model. In summary: rolling window = new observations are added to the training database. (fixed starting point, sample size increases) moving window = add new observation, same time drops out the oldest observation. (fixed sample size). The question is How Matlab NEWELM function works with these different approaches??? regards JP
|
Pages: 1 Prev: duality gap value in linprog Next: Read Image to Matrix in Simulink |