Loading ...
Sorry, an error occurred while loading the content.

Re: [GP] Looking for GPLab Tutorial for newbies

Expand Messages
  • Sara Silva
    Hi Bill, Assuming that everything I said before is correct (that time series prediction can be done like symbolic regression), your files are fine. Now all you
    Message 1 of 4 , Nov 1, 2006
    • 0 Attachment
      Hi Bill,

      Assuming that everything I said before is correct (that time series
      prediction can be done like symbolic regression), your files are fine.
      Now all you need to do is to slightly change the demo.m file of GPLAB so
      you can use these files (which should be in the GPLAB directory):

      p.datafilex='name_of_your_x_file';
      p.datafiley='name_of_your_y_file';

      Also, you do not want to use test data:

      p.usetestdata=0;

      You may want to change other parameters, for example make the runs
      faster by not calculating introns and other complexity measures, as well
      as diversity measures:

      p.calccomplexity=0;
      p.calcdiversity=0;

      You can also eliminate all plots during the run:

      p.graphics={};

      There are many other parameters you may change. Please refer to the
      user's manual for details. I hope this helps.

      Cheers,
      Sara



      Bill Ngu wrote:

      > Hi Sara,
      >
      > Thanks for your comment. Can I confirm that I need to prepare 2
      > file, ie one y file and one x file?
      >
      > I am using the following stock market data:
      >
      > 1088.52
      > 1084.55
      > 1072.49
      > 1041.27
      > 1041.91
      > 1041.39
      > 1058.62
      > 1071.03
      > 1078.20
      > 1081.19
      > 1081.88
      >
      > If I want to use to use 5 points to predict the next one... Refering
      > to the parity3bit xy file and your explaination, I prepare my X file
      > like below:-
      >
      > 1088.52 1084.55 1072.49 1041.27 1041.91
      > 1084.55 1072.49 1041.27 1041.91 1041.39
      > 1072.49 1041.27 1041.91 1041.39 1058.62
      > 1041.27 1041.91 1041.39 1058.62 1071.03
      > 1041.91 1041.39 1041.39 1058.62 1071.03
      > 1041.39 1058.62 1071.03 1078.20 1081.19
      > 1058.62 1071.03 1078.20 1081.19 1081.88
      >
      > And following is my Y file as below:-
      >
      > 1041.39
      > 1058.62
      > 1071.03
      > 1078.20
      > 1081.19
      > 1081.88
      >
      > Can you confirm that what I do is correct for time series prediction?
      >
      > If the above X Y files is in correct format, what should I do next?
      > Thanks
      >
      > Regards,
      >
      > Bill
      >
      >
      > --- In genetic_programming@yahoogroups.com, Sara Silva <sara@...>
      > wrote:
      >
      >>Hi there,
      >>
      >>As far as I know, there is no tutorial for GPLAB except the manual
      >>itself. GPLAB does not have a time series module, but I suppose it
      >
      > would
      >
      >>not be too difficult to use it for that purpose. I have never done
      >
      > time
      >
      >>series prediction myself, so all you experts out there please let
      >
      > me
      >
      >>know if I say something really stupid.
      >>
      >>Time series prediction can be handled like symbolic regression
      >
      > would. If
      >
      >>you have a series like
      >>
      >>0
      >>1
      >>2
      >>3
      >>4
      >>5
      >>6
      >>7
      >>8
      >>9
      >>
      >>and you want to use 5 points to predict the next one, you should
      >
      > have
      >
      >>your x and y data files like this:
      >>
      >>x y
      >>0 1 2 3 4 5
      >>1 2 3 4 5 6
      >>2 3 4 5 6 7
      >>... ...
      >>
      >>So, probably all you need is a data importation function to
      >
      > transform
      >
      >>your data series as described, and then you can use GPLAB as you
      >
      > would
      >
      >>for symbolic regression. This may not be the most efficient wat to
      >
      > do
      >
      >>it, but it certainly is a simple solution. Once again, please let
      >
      > me
      >
      >>know if this doesn't make sense.
      >>
      >>GPLAB is really biased towards my own research, but I have had so
      >
      > many
      >
      >>users asking me how to do time series prediction that I will
      >
      > eventually
      >
      >>add something for that purpose. But don't hold your breath!
      >>
      >>Cheers,
      >>Sara
      >>
      >>
      >>
      >>Bill Ngu wrote:
      >>
      >>
      >>>Hi All,
      >>>
      >>>I am a master student doing research on stock market prediction
      >
      > using
      >
      >>>Genetic Programming. I had downloaded GPLAB 2 for MATHLAB.
      >>>
      >>>However, it give me a hard time to figure out how does this time
      >
      > work
      >
      >>>and how can I use it to predict a financial time series.
      >>>
      >>>May I know if anyone can help to point me to a GPLab tutorial
      >
      > for a
      >
      >>>newbies? Anyone can shed some light on how should I use GPLab
      >
      > toolbox
      >
      >>>for financial time series prediction?
      >>>
      >>>Thanks
      >>>
      >>>
      >>>
      >>>
      >>>
      >>>
      >>>
      >>>
      >>>
      >>>Yahoo! Groups Links
      >>>
      >>>
      >>>
      >>>
      >>>
      >>>
      >>>
      >>>
      >>>
      >>>
      >>
      >
      >
      >
      >
      >
      >
      >
      > Yahoo! Groups Links
      >
      >
      >
      >
      >
    Your message has been successfully submitted and would be delivered to recipients shortly.