Machine learning in trading: theory, models, practice and algo-trading - page 357
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
You may see a cyclic pattern.
Decompose it in Fourier (in R it is 2-3 lines), plot it, and you won't see any cyclicity. It is an even spectrum.
Try the autocorrelation function. Silence again, but it should come out if there is cyclicity.
Decompose it in Fourier (in R it is 2-3 lines), plot it, and you won't see any cyclicity. It is an even spectrum.
Try the autocorrelation function. Silence again, but it should come out if there is cyclic frequency.
I think to teach NS at once, feed several different periods with these histograms, it will consider both larger and smaller changes, large will be a trend component and small will be signals to enter
And then convert the signals taking into account the current charts and their relation to bollinger bars. The method of scientific experimentation, in short
With the period of the boyles 1 it will be like this :)
And you can see the cyclicality
Stationary row. And the emissions are removed - standard procedure
try to take the emissions into account too, LSTM, in theory, should cope
I think to teach the NS at once, feed several different periods with these histograms, will take into account both larger changes and smaller, large will be a trend component and small signals to enter
And then convert the signals taking into account the current charts and their relation to bollinger. So, let's try it scientifically.
Does anyone know the answer to the question: how do NSs treat non-stationary entrances?
bad? )
Does anyone know the answer to the question: how do NS refer to non-stationary inputs?
I've been studying the method I would have simulated it to see if there is anything worthwhile in it. Then the training will be more adequate.
Everything is clear here, extrema signal a reversal, especially the short ones, an antipersistent series is present - a new peak is followed by a new trough (in the indicator with a small period)
The opposite situation is in the indicator with a long period, the series looks persistent, a new maximum is followed by a new maximum, i.e. you can identify the long trends, at the same time, the series is stationary and you can find (approximately) the end of the trends
I've read a lot of books, haven't I?
Working algorithm: we define the trend and work by it, carrying out entries on the anti-peristence series, to increase the probability of winning. At the same time, if the indicator with a large period is close to the mean extremums, we change the deal entries to the opposite ones, if the trend starts to changeWell, everything is clear here, extrema signal a reversal, especially of short periods, an antipersistent series is present - a new peak is followed by a new trough (on the indicator with a small period)
The opposite situation on the indicator with a long period, the series looks persistent, a new maximum is followed by a new maximum, i.e. you can identify the long trends, at the same time the series is stationary and you can find (approximately) the end of the trends
I've read a lot of books, eh?
Cool.) What's the antipersistent?
A new maximum (probably minimum) is followed by a new maximum - yes, I've seen it too, the graphs are all familiar. You simulate - and there's nothing there - empty. Maybe you'll be lucky.