Machine learning in trading: theory, models, practice and algo-trading - page 2290

[Maxim Dmitrievsky]

Vitaly Muzichenko:

Well, it's hard for me to even imagine where to start and what it should look like. In my mind, these are kind of incompatible things.

It makes more sense than it might seem at first glance.

[Renat Akhtyamov]

Vitaly Muzichenko:

Well, it's hard for me to even imagine where to start and what it should look like. In my mind, this is kind of incompatible things.

First, an overview of existing systems. Tests. Comparative characteristics, pros and cons. Fixing drawbacks, ideas. Shaping TS. What do you observe the need to involve mo?

[Aleksey Mavrin]

Maxim Dmitrievsky:

It makes more sense than it might seem at first glance

What do you mean - what is the role of a neural network - to find moments for the first entry? Or to teach the network to find the sense of opening with increased lot in minus?)

[Maxim Dmitrievsky]

Aleksey Mavrin:

What do you mean - what is the role of the neural network - to find moments for the first entry? Or to teach the network to find the sense of opening with a higher lot in minus?)

I'm interested in practical implementations, success stories

I saw one in Signals, works fine

[Vladimir Perervenko]

Renat Fatkhullin:
You can share the information:

1) Do you use the python library of MT5?

2) Do you use it outside or inside MT5?

3) What functions does the library lack? Access to indicators?

We are preparing an upgrade of MQL5 adding fast matrix operations. This will allow to perform massive calculations normally.

We will also develop connectors to analytical packages and implement the standard WinML integration.

I will use the library.

2. If you mean do I use Meta Editors for creating and debugging Python scripts - no.

I don't.

• Receiving the events of the terminal in Py scripts.
• exchange of data with MKL program
• Indicators? desirable but not necessary. Any indicator can be calculated in R/Py

The plans are impressive. But it seems to me that you want to grasp the immensity.

Good luck

PS: Do you still think that discussion of R language usage is inappropriate in this forum?

[Renat Fatkhullin]

Vladimir Perervenko:

1. use the library .

2. If you mean using Meta Editors to create and debug Python scripts, no.

I meant launching Python scripts in the terminal as scripts.

3. not enough.

• Receiving terminal events in Py scripts.
• exchange of data with MKL program
• Indicators? desirable but not necessary. Any indicator can be calculated in R/Py

Events will not be available as it is a scripting mode.

The plans are impressive. But it seems to me that you want to embrace the immensity.

Look around the MetaQuotes & MetaTrader & MQL ecosystem, it was all immense too. And it's always been heard "you can't".

Plus, you see/understand very little of it.

PS: Do you still think it's inappropriate to discuss R language application in this forum?

Discuss, why not. But don't ask us to do something for R. The subject is closed for us.

[Vladimir Perervenko]

Renat Fatkhullin:

I meant to run python scripts in the terminal as scripts.

Events will not be there, as it is a script launching mode.

Look around the MetaQuotes & MetaTrader & MQL ecosystem, it was all immense too. And it was always heard "you can't".

Plus, you see/understand very little of it.

Discuss why not. But don't ask us to do something for R. The issue is closed for us.

Running python scripts in the terminal as scripts I use.

We will not ask to make something for R, what we have is enough for work.

Good luck

[Rorschach]

Rorschach:

Ideally, you should make the series stationary, find cycles on the spectrum, build a filter for these cycles.

How to make the series stationary? Correction for the average volatility by the time of day? Logarithm and filtering?

How do you build a spectrum? The deeper the history is, the stronger the noise is, the weaker the signal can be. But everything changes on the market. If we take a short time span, we will not get a random coincidence but a cycle.

I've shown my statistics on the spectrum

I can say that the problem is solved. Time does not carry any information and I'll leave it alone. I will also calculate persistence, but most likely there will be nothing new.

Question 2, we can proceed from the required quality, for example, for an accuracy of 95% and a sample of 400 bars, the error will be +-0.1*sco. Or signal/noise ratio. If you add up 100 sine waves, the amplitude will increase 100 times, and if you add up 100 realizations of noise, only 10 times.

[Maxim Dmitrievsky]

Rorschach:

1 the question can be said to be solved. Time does not carry information, take the equitibles bars and do not bother. I will also calculate persistence, but most likely there will be nothing new.

Question 2, we can proceed from the required quality, for example, for an accuracy of 95% and a sample of 400 bars, the error will be +-0.1*sco. Or signal/noise ratio. If you add up 100 sine waves, the amplitude will increase 100 times, and if you add up 100 realizations of noise, only 10 times.

my research shows the opposite picture

[Maxim Dmitrievsky]

articles are beneficial to the minds of people