Given a set of at least 180 data-points (dates as X and associated non-negative values as Y) - how would one go about predicting/forecasting future data-points?

I'm not even sure whether what I'm looking for is 'non-linear extrapolation', but searching on things like newton/lagrange, polynomial regression, etc - I just don't seem to be able to find one formula, but perhaps that's because it can't be simplified like that. I also read that it depends a lot on the data.

My data-points are prices that go up and down, and follow no mathematical equation, not random - and frequency, oscillation and so forth, are unknown. Is this possible to achieve? I'm not interested in using 3rd-party programs like Excel/Matlab, but rather develop the model. Thanks in advance.

  • $\begingroup$ It is a very general problem which can have many different solutions depending on lots of stuff. So it's kind of difficult giving only one approach. $\endgroup$ – mathreadler Aug 17 '16 at 10:26
  • $\begingroup$ I see.. Is there a specific topic i should be researching on, since I'm still not exactly sure which topic covers what I'm trying to do? Can you point me in a direction? :-) $\endgroup$ – Jeppe Aug 17 '16 at 10:32
  • $\begingroup$ I suppose that your speak about time series. There are specific techniques for them. Have a look at itl.nist.gov/div898/handbook/pmc/section4/pmc4.htm $\endgroup$ – Claude Leibovici Aug 17 '16 at 10:43
  • $\begingroup$ There are so many places to point I don't know where to start :) $\endgroup$ – mathreadler Aug 17 '16 at 10:48
  • $\begingroup$ Maybe it will be easier to propose a method if we know more about the data which is to be extrapolated. $\endgroup$ – mathreadler Aug 20 '16 at 20:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.