Fast Fourier Transform - Cycle Extraction - page 6
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Try this indicator with SSA: ?????????? ??????? ???????????? ?????????????? - MQL4 Code Base
I already tried,I have found it to be non reliable for prediction and pc resource consuming.
My question was related to the suggested implementation of COMBINED SSA+MESA analysis as per the manual of UCLA software,since you seemed to have everything up and running,it seemed to me that you could just apply MESA to your saved reconstructed SSA series...Up to you if you want to post,I will do it if/when I succeed in installing the software.
Goertzel
Note about Goertzel:
I think that we have been misreading Goertzel indicator..instead of using ,for example,0.050 as 5% of 720 periods(36 bars cycle)...I believe that the right reading is 0.050 periods per bar(20 bars cycle)...I have done a test with values 360,540 and 720 and the results achieved have led me to believe this is the right interpretation..Anybody can confirm/deny?
mtm data series
Where it is the MESA software?
mtm data series Where it is the MESA software?
MEM Spectrum mode...apply it to reconstructed SSA,in theory this will give us very good cycle frequencies.
SIMBA, could you give an example for refining opnion on ue of the Goertzel? This may not fit in line with what you mean, but what about using 18 instead of 20? Not a large difference but perhaps more accurate?
I think what barnix is meaning is to apply those MTM values to the SSA plugin.
I will load the toolkit when I get some time, and if successful will post steps for use in a win32 environment.
Codobro&Barnix
SIMBA, could you give an example for refining opnion on ue of the Goertzel? This may not fit in line with what you mean, but what about using 18 instead of 20? Not a large difference but perhaps more accurate?
I think what barnix is meaning is to apply those MTM values to the SSA plugin.
I will load the toolkit when I get some time, and if successful will post steps for use in a win32 environment.Codobro,Barnix:
Thanks for your help,see if I can install it by myself.
1-Regarding the toolkit,I read all the explanations of the instruction manual,plus the comments at their website,seems to me that a very reliable method is to reconstruct a time series by using SSA(you increase signal to noise ratio),then apply MESA(MEM in the toolkit) to the reconstructed SSA time series(because MEM works exceptionally well with high signal to noise ratio time series) and extract a reliable cyclic model...One that we can use for trading GBPUSD D1.
2-Additionally,when looking at how the UCLA people explained the MTM,they pay special attention to the "black bars above 99% confidence level"..so,these are the ones I am asking barnix about...I would like to know,exactly,at what frequencies are they located,since,just by visually examining his chart I can only see approximate frequencies...If I knew the right frequencies,I could create a cyclic composite for GBPUSD,post it here ,and use it for testing or trading GBPUSD D1...until I don`t get a reply from barnix,what he has shown is very interesting,but we can`t use it for trading.
3-Using the "Full SSA" indicator,without further analysis, is a waste of time,I had been testing it for a couple of weeks with no results(and very heavy on cpu) on several timeframes,yes,it looks nice but predictive value is zero..And this is logical since we don`t know how many components to use,nor the lag...to know these we need to check the eigenvectors with the MTM toolkit.
4-Codobro,regarding Goertzel I don`t know what else to think...According to cycle theory(which I don`t believe 100% )lower frequencies should have higher amplitudes and this is coincident with % of 720 bars(or whatever n you use)...but when using 720,540 and 360 periods..very similar numbers,for highest amplitude cycle,appear..like 0.0517,0.0522...etc and this has led me to understand that for this to be right the meaning of the number must be taken at 0.0517 or 0.0522 cycles per period,so,that when you analyze 720 bars or when you analyze 540 bars or when you analyze 360 bars,the same cycle is still there(approx 19 or 20 periods)...AND...if you stop for a moment and visualize it,this means that the higher amplitudes are achieved with the higher frequency/lower period cycles(0.05 means 20 periods...0.01 means 100 periods)...So,I am lost and looking at the indicator code didn`t help me at all
The MTM frequencies and reconstruction with 99% frequencies (mtmspec-sign.out from
mtm1.zip):
0.00097656 99%\0
0.00390625 99%\0
0.00927734 99%\0
0.01171875 99%\0
0.01660156 99%\0
0.02246094 99%\0
0.02490234 95%\0
0.02929688 99%\0
0.03613281 99%\0
0.03857422 95%\0
0.04785156 95%\0
0.05761719 90%\0
0.06787109 90%\0
0.07763672 99%\0
0.09179688 95%\0
0.10107422 95%\0
0.11962891 95%\0
0.12353516 99%\0
0.12597656 90%\0
0.14501953 95%\0
0.15185547 90%\0
0.15869141 95%\0
0.16406250 95%\0
0.22656250 95%\0
0.23339844 99%\0
0.23925781 90%\0
0.30029297 99%\0
0.30957031 95%\0
0.31445312 95%\0
0.34082031 99%\0
0.35156250 95%\0
0.35498047 95%\0
0.37207031 99%\0
0.39062500 99%\0
0.40380859 99%\0
0.41357422 95%\0
0.42089844 95%\0
0.45703125 99%\0
0.47021484 95%\0
0.49951172 99%\0
Barnix,these are the interesting ones
The MTM frequencies and reconstruction with 99% frequencies (mtmspec-sign.out from
mtm1.zip):
0.00097656 99%\0
0.00390625 99%\0
0.00927734 99%\0
0.01171875 99%\0
0.01660156 99%\0
0.02246094 99%\0
0.02490234 95%\0
0.02929688 99%\0 ******
0.03613281 99%\0******
0.03857422 95%\0
0.04785156 95%\0
0.05761719 90%\0
0.06787109 90%\0
0.07763672 99%\0
0.09179688 95%\0
0.10107422 95%\0
0.11962891 95%\0
0.12353516 99%\0*****
0.12597656 90%\0
0.14501953 95%\0
0.15185547 90%\0
0.15869141 95%\0
0.16406250 95%\0
0.22656250 95%\0
0.23339844 99%\0
0.23925781 90%\0
0.30029297 99%\0
0.30957031 95%\0
0.31445312 95%\0
0.34082031 99%\0
0.35156250 95%\0
0.35498047 95%\0
0.37207031 99%\0
0.39062500 99%\0
0.40380859 99%\0
0.41357422 95%\0
0.42089844 95%\0
0.45703125 99%\0
0.47021484 95%\0
0.49951172 99%\0**********Barnix,
Thanks a lot,
The ones I marked with asterisks are extremely interesting for avery simple noise free reconstruction of GBPUSD D1.
SIMBA, that is an interesting way to look at Goertzel, thank you, I shall try it. I am also very interested in looking over the composite you spoke of when its built
Thank you barnix for listing the values as well, to my understanding the marked ones correlate to values given by the digital filters.
SIMBA, that is an interesting way to look at Goertzel, thank you, I shall try it. I am also very interested in looking over the composite you spoke of when its built
Codobro,
while I try to build a composite that doesn`t repaint,you can build one yourself by using any of the indicators barnix posted at this thread...just use these periods for GBPUSD D1 :
2,8,28,34...so freq min/max are 1/3,8/9,27/28,34/35...will try to attach the indicator here...found,you will need the fft libraries that are here in this thread.