Dear All,
I am working on time series forecasting using neural networks. I have used
the neural network toolbox in the past for other applications (pattern
classification) but this time I face difficulties on how to present my
data set to MATLAB.
Since I work on time series forecasting, I have three data sets consisting
of 100 indices (prices of a commodity during a period of 100 months). I
expect the NN to perform two tasks :
1st. To predict prices of the commodity based on actual past values. For
example, in a feed forward NN, supplying the two first values I expect
from the network to predict the third, supplying the 1st, 2nd and 3rd, to
predict the 4th, and so on. (One-lag)
2nd. To predict prices, when I append previously predicted values to the
input database, and use these values to predict future values of the
commodity. For example, if the network is used to predict a value n6 from
observed input data i1,i2,…,i5 then the next network prediction n7 is made
using inputs i2,i3,i4,i5,n6. (Multi-lag)
My question is on how I should enter the data in MATLAB in order the
network to perform these two tasks.
Regards,
Constantinos Alexandrides
|