Machine learning in trading: theory, models, practice and algo-trading - page 1730

New comment
 
Maxim Dmitrievsky:

Tried 10 years ago caterpillar, fish are not there in the classic form

Well, you're weird, caterpillar does not work because it is current and previous and previous redraws, about what fish you're talking about I do not understand. Roughly speaking on a new bar CSSA and SSA will show different values and it is SSA in the process of building the bar will change towards CSSA, which will stand as in the copan. Do not confuse the coast, Max, do you think I would pay attention to a specific fuflomitsin and even so respectable people to distract? But this approach is absolutely workable in terms of redrawing and technically, another thing is that I was able to see it only once. That's what I wanted to check. You're all frequency engineers are crazy about frequencies, so check the strategy, get busy. CSSA method was actually published later, the public was eager to know how so not redrawing caterpillar, which was the only achilles' heel, which does not allow it to use, and it beckons with its perfect shapes on the history....
 
mytarmailS:

...

Thank you for the benefit...

Did you write the program yourself?

The decomposition itself is ready, and the rest is my matlab script based on examples from the tutorial

 
mytarmailS:

I don't get it... Explain in more detail, what is the root and where did I go wrong?

We are looking for discrete states that depend either on time or on indicator readings or whatever else. The moments of increments in these states remain constant for a long time, it remains to write a simple TS for them

This is what you started to do with temporal glues, but you haven't developed the topic, you haven't generalized it to other cases

you can use clustering, but it's harder to interpret
 
Mihail Marchukajtes:

is there anything/where to read about the CSSA?

 
Vladimir Suslov:

the decomposition itself is ready, and the rest is my matlab script by examples from the tutorial

You're cute!

I use the "Rssa" package from R

https://arxiv.org/pdf/1309.5050.pdf

http://www.milanor.net/blog/wp-content/uploads/2014/07/SingularSpectrumAnalysisWithRssa.pdf

 
mytarmailS:

Max the point is that this state today is the size of 5 candles, tomorrow 3 candles after tomorrow 23 candles all as in life, everything flows all changes and will never be the same as before, will be similar but you so ...

And when you train AMO on the states of 5,3,23 candles, the next one will be 203! Do you understand?

Spectral analysis should help, arrange everything in the order, look what changes, increases, decreases, etc.

And then, with an understanding of the process it is possible to build something adaptive.

riposte

 
Vladimir Suslov:

is there anything/where to read about CSSA?

The documentation is official, but they are bourgeois and everything was difficult to understand. NOXA ANALITYCS
 
Mihail Marchukajtes:

If at that moment the price went like that, then the indicator could work very well, the period is almost stationary.


 
Maxim Dmitrievsky:

peep

)))) I already have a caterpillar in my head)))

Maxim Dmitrievsky:

We are looking for discrete states that depend either on time or on indicator readings or whatever. The moments of increments in these states remain constant for a long time, all that remains is to write a simple TS for them

This is what you've started to do with temporal glues but hasn't generalized to other cases

you can use clustering, but it is more difficult to interpret
You can also clustering, also tried , you can also take two or three clusters, but there is already "curse of dimensionality" few examples
 
mytarmailS:

)))

I already have a caterpillar in my head)))

the caterpillar can be used BEFORE the pretreatment, then you can cover yourself in Nutella and pour champagne

1...172317241725172617271728172917301731173217331734173517361737...3699
New comment