Machine learning in trading: theory, models, practice and algo-trading - page 1879

 
mytarmailS:

1) because it predicts both the current hour (past) and the next hour (future)

which is not good because we already know half of the prediction

2) three clusters generalize the situation too roughly...

as you can see from the picture, it is a buy cluster, because the general movement was up, but it was at the first hour, and we trade the second

so we trade the second hour as buy, but it's actually a sell.

And when we predict the suto cluster of the next hour without 0.5 :)

Not a fact. Maybe there are some patterns of sell coaster in buy cluster that NS should find... That's what it's all about. Do you find it?

 
Mihail Marchukajtes:

Not a fact. Maybe there are patterns in the buy cluster of the sell coaster, which the NS should reveal... That's what it's all about. Do you find it?

No, I do not.

 
Who has seen a working neural network on forex? Well, so that it worked according to a known algorithm and the results of real trading could be seen? Maybe someone is already trading on it? I do not question the efficiency of neural networks. I would like to compare the results.
 
mytarmailS:

No, I don't.

How do you do it? With a look?
 
Ivan_Invanov:
Has anyone seen a working neural network on forex? Well, so it worked according to a known algorithm and the results of real trading could be seen? Maybe someone is already trading on it? I do not question the efficiency of neural networks. I would like to compare the results.

For example, I traded not on forex, but on the real market. What next? Any suggestions for intervention?

 

It is probably impossible to explain more accessible about Ns, at least I have not met.

And the article

 
Maxim Dmitrievsky:
In short, I will do it.)

Don't go too far, don't go too high))))

 
Ivan_Invanov:
Has anyone seen a working neural network on forex? Well, so it worked according to a known algorithm and the results of real trading could be seen? Maybe someone is already trading on it? I do not question the efficiency of neural networks. I would like to compare the results.
Why was my post deleted?
the man asked for a working neural network on forex, I gave a link to my project, which is "a working neural network on forex,
what is the problem?
 
Evgeny Dyuka:
why did you delete my message?
the man asked a working neural network on forex, I gave a link to my project which is "working neural network on forex",
what is the problem?
ay, admin, this is not a rhetorical question, please answer,
for example: "our rules prohibit..."
to know
 
mytarmailS:

1) because it predicts both the current hour (past) and the next hour (future)

which is not good because we already know half of the forecast

2) three clusters generalize the situation too roughly...

as you can see from the picture, it is a buy cluster, because the general movement was up, but it was at the first hour, and we trade the second

so we trade the second hour as buy, but it's actually a sell.

And when we predict the suto cluster of the next hour without the previous acurasi 0.5 :)

so you normalize the 2nd to the 0-th bar, then it won't be 0.5

There is a linear change there, i.e. the curves lie on 1 line. So the correlation between clocks tends to one. Check it... use your brain

3 is not enough - make it 500 :) but justify why