Machine learning in trading: theory, models, practice and algo-trading - page 1974

 
NeuralNetwork:

Hello Maxim Dmitrievsky. I just read your email. There's nothing on githab. I took a simple network code and modified it to suit me. What was decisive for me was removing the garbage from the tick history. Python

Hmmm... i don't really get it... so you removed all repetitive prices... i don't know what that gives

 
Maxim Dmitrievsky:

Hmmm... I don't really understand... i.e. you removed all the repeating prices... it is unclear what this gives

I wanted to find the top and bottom for a certain number of ticks, so that the price is guided by them when training the neural network.

 
You understand what this is for? When we're in a flat, then the most important thing is not to lose money. And the trend is easily recognized by the network. The most important thing for a neural network after guessing the trend direction during a flat.
 
NeuralNetwork:
You understand what this is for? When we're in a flat, then the most important thing is not to lose money. And the trend is easily recognized by the network. The most important thing for a neural network is to guess the direction of the trend during a flat.

Well... it depends on what you teach. What to enter and what to exit

 
NeuralNetwork:
You understand what this is for? When we're in a flat, then the most important thing is not to lose money. And the trend is easily recognized by the network. The most important thing for a neural network is to guess the trend direction during a flat.

Why not record the accumulation of ticks in a selected window?

 
Aleksey Vyazmikin:

Why not record the accumulation of ticks for the selected window?

If that's what you mean I'm fixing the number of ticks

df = df[-240000::1]
 
NeuralNetwork:

If that's what you mean, I already fix the number of ticks

I meant that it would rather aggregate the price of ticks in a flat, for example with a negative value, as soon as you understood what is in the flat in the form of a separate predicator. If you removed repetitive prices, then the predictors were removed with them, or did you build them on cleared ticks?

 
Aleksey Vyazmikin:

I rather meant that it would aggregate the price of ticks in the flat, for example by a negative value, as soon as you understood what is in the flat in the form of a separate predicator. If you remove repetitive prices, then predictors are removed with them, or you construct them using already cleared ticks?

My network is set up in such a way that I may build on cleared ones.

 
NeuralNetwork:

My network is set up so that I build on cleared ones. You should try it.

And what kind of predictors go to the input - increments?

 
mytarmailS:

I've tested it with spread - well, it seems to drag, it's worse. I will rewrite it for the terminal next week and it will be possible to put it for tests.