Using Fourier Transform for Forecast?


Using the trigonometric representation (i.e. sum of cosines) of discrete Fourier transform (DFT), can I apply for forecasting new values? 


1 comment
  • No, using DFT trigonometric representation for forecast is not valid.

    Why? The DFT assumes that your data set is periodic, i.e. it repeats itself every N samples, so if you were to use it for forecast, the value you get for N+1 (using full spectrum) would be exactly the same value of observation at t=0.

    When can you use it? Use it when you wish to interpolate values for intermediate points.

    More information? see the following link for more details

    Comment actions Permalink

Please sign in to leave a comment.

Didn't find what you were looking for?

New post