Machine learning in trading: theory, models, practice and algo-trading - page 2485

[Renat Akhtyamov]

mytarmailS #:

buy the eu for half a day minimum.

humor section : signal from me) )

any luck with the pass?

[mytarmailS]

Renat Akhtyamov #:
were you able to search?

No, I haven't found where these values come from

[Renat Akhtyamov]

mytarmailS #:
No, I never found where these values come from

That's weird.

Send me a link in your email, I'll try it tomorrow.

[mytarmailS]

Renat Akhtyamov #:

That's weird.

Send me the link in your personal message, I'll try it tomorrow.

So the link to the site on the previous page, you opened it yourself, or I do not understand... or you do not open the demo?

[Renat Akhtyamov]

mytarmailS #:
So on the last page link site, you opened yourself, or I do not understand... or you do not open the demo?

No, yandex.

[Mihail Marchukajtes]

JeeyCi #:

please answer my clinical question as well (you, by the way, read my thoughts yesterday and after I have already looked at this method, posted your way of working with data - thanks)... BUT the question remains: this method is used to classify, I guess, features - do you need it... what do you classify, if it's not a secret? LN(Close/Open)? and what do you teach?

-- if it's a secret, i understand - "know-how"?

p.s. I'll throw myself a couple of links for orientation in the topic (after all, it's not really my statistics, although the latter can be put into an "environmental model", probably):

Introduction to AI

Statement and possible solutions to the problem of training neural networks

preprocessing

An ensemble of methods

Yes indeed it is a method of classification, where I teach the model to recognize the truth or falsity of the signals of the basic strategy. That is, when working with classifications in the composition of the TS should be a basic strategy. The strategy itself may be absolutely any, the same crossing of bars and all models have 50/50 correct and incorrect signals. The task of classification is to determine which signals are really true and which are false. Read my article, it is described in more detail there!

[JeeyCi]

Mihail Marchukajtes #:
Read my article there for more details!

thanks for the link and the article... If the data is based on ClucterDelta - that's an encouraging start... but Spot does not always walk like Futures (as far as forex is concerned)...

BUT the conclusion about the truth/falsehood of the signal, as far as I understand, is still based on Bayes...?

By the way, here(p.20) is a crash of my attempts to figure the NS graph (having input option price distributions):

Bayesian inference differs from traditional statistical inference in that it preserves uncertainty . ..

The Bayesian worldview interprets probability as a measure of the likelihood of an event , that is, the degree to which we are confident that the event will occur.

... although its parameters (the existing distribution) can also be tried to be fed into the input, probably - probably then look towards multiclass classification

[Mihail Marchukajtes]

JeeyCi #:

Thanks for the link and the article... If ClucterDelta data is the basis, it's a reassuring start... only Spot does not always run like Futures (if we talk about forex)...

BUT the conclusion about the truth/falsehood of the signal, as far as I understand, is still based on Bayes...?

By the way, here(c.20) is the collapse of my attempts to figure the NS graph (having input option price distributions):

... although its parameters (the available distribution) can also be tried to feed into the input, probably - probably then look towards multiclass classification

At the moment I don't use ClusterDelta anymore since I switched to Moex and there this information is free plus there is also information on OI, but as for the options, you need to input the values of smile parameters which are 3. The curvature, the slope and the value in the central strike and not the values themselves but their change over time. This is what I still do not have, alas, and then the strategy will be almost a win-win!!!!! It seems to me....

[JeeyCi]

Mihail Marchukajtes

I got up the strength/courage to look through your code (often there is more truth in the code than in all textbooks) - can you tell me what are those multipliers in your Classifiers in the variable double decision - are they weights?... and how did you initially find them? i.e. why exactly those?

or better yet, comment please - what variables does it take, and the function code

double getBinaryClassificator1(double v0,double v1,double v2,double v3) 
  {
   double x0=2.0 *(v0+327.0)/650.0-1.0;
//Variable v1 got under reduction
   double x2 = 2.0 * (v2 + 397.0) / 828.0 - 1.0;
   double x3 = 2.0 * (v3 + 121.0) / 264.0 - 1.0;
   double decision=1.5260326743246075*x0
                   -0.13861638107404117 * sigmoid(x0)
                   -0.06391652777916389 * sigmoid(x2)
                   -0.44591870340615364 * sigmoid(x0 + x2)
                   +0.14661031327032664*sigmoid(x3)
                   -0.024191375335575492*sigmoid(x0+x3);
   return decision;
  }

thanks in advance!

p.s.

1. I see that you use a sigmoid (S-shaped) function as an activation function... it is "often used as a compressing function"...

2.

Mihail Marchukajtes #:
... and not the values themselves, but their changes over time.

maybe it would be better to squared?

[JeeyCi]

By the way, volatility is volatility (as a non-systemic risk), but no one cancelled the systemic risk...

Volatility in financial markets is not the same as risk

p.s.

Although, of course, a trader earns on volatility... imho