I would like to build trading signals using cyclic analysis in order to obtain a forecast afterwards. I had a look in literature and Hurst analysis, Fourier, etc, are used

However, I am struggling to capture what really is the state of the art at moment. As I am not coming from finance but from engineering, I was more willing to use wavelet analysis as they better fit the non-stationary behaviour of a financial price.

In principle one can still use Fourier, but you need to detrend the data afterwards and you lose anyway the temporal scale that you instead have in wavelet (or in STFT).

Can anyone point me to the right direction on the state of the art? In case, you think cyclic analysis is not a good idea to generate trading signals, please elaborate the reasons why is that.

  • 2
    $\begingroup$ There is an interesting post on cyclic analysis (of the Fourier type) here quant.stackexchange.com/questions/61565/… Notice in the chart, that the fit is excellent in-sample, but the forecast and the actual completely diverge in the out-of sample period. Something to keep in mind when you plan to do forecasting using such a method. $\endgroup$ – noob2 Apr 14 at 9:10
  • $\begingroup$ thanks for the link. The out of sample is really bad indeed even if I am not convinced the code is entirely correct (both of them Q&A) $\endgroup$ – Luigi87 Apr 14 at 9:25
  • $\begingroup$ If you try to work with prices as opposed to returns, you will see this problem in post cases. Any kind of fit to historical data (whether Fourier, wavelet or whatever) is most often bound to fail as prices are most often not stationary 'enough'. Notwithstanding my comment, it's a nice little project, though. $\endgroup$ – Kermittfrog Apr 14 at 9:47
  • $\begingroup$ So although the wavelet are indeed use to dealt with non stationary signals, including shock phenomenon, it returns coefficients from which future can be estimated in a deterministc way. This does not include though the fact that prices are by nature non-deterministic so I probably agree then that in-sample they should perform great but poorly out of sample.. $\endgroup$ – Luigi87 Apr 14 at 11:27

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.