Prev: Searching cell array
Next: Plot gaussian contour
From: Jeff Spenser on 8 Feb 2010 18:32
Hi

Im having some problems calculating predictions for ARMAX model which I created. The model is modeling rougly 100 datapoints and I which to predict the next hundred or so which are unknown now.
The input variables are known for the 100 datapoints which are also known and also for the next 100. What I have done is creating a model for those 100 datapoints where the first 100 points of the input variables are used also using System Identification Toolbox

Exported that model to the workspace and then did a predict by:

model = model exported from ident
input_predict = [input_first_100; zeros(100,1)];
output = [[var1_first_100; var1_last_100] [var2_first_100; var2_last_100] [var3_first_100; va3r_last_100]];

data = iddata(input_predict,output);
predict(model,data)

Is this methood correct? What I am wondering about is how to get matlab to realise that the input variable is unknown rather than simply 0 for datapoints 101-200. Or should this only be a predict for the future, that is:

input_predict = zeros(100,1)
output = [var1_last_100 var2_last_100 var3_last_100];

Also - is it possible to make matlab plot with the prediction output some confidence interval, for exapmle 95% confidece, based on the model specifications?
 | 
Pages: 1
Prev: Searching cell array
Next: Plot gaussian contour