Simple nonlinear prediction

predict -d# -m# [-r# | -v#] [-s# -o outfile -l# -x# -c# -V# -h] file(s)

-d delay
-m embedding dimension
-r absolute radius of neighbourhoods
-v same as fraction of standard deviation
-s time steps ahead forecast (one step)
-l number of values to be read (all)
-x number of values to be skipped (0)
-c column to be read (1 or file,#)
-o output file name, just -o means file_pred
-V verbosity level (0 = only fatal errors)
-h show this message

Performs locally constant predictions on scalar time series and prints the root mean squared prediction error. Predictions are made a time ahead given by -s. Either -r or -v must be present. Predictions are written to file_pred.

Note: A version of fclazy is used that implements fast neighbour search (not like in the book).

See also zeroth which does essentially the same but has some different options, including multivariate data, and xzero, which does cross-predictions.

Table of Contents * TISEAN home