Discussion of article "Practical application of neural networks in trading"

[MetaQuotes]

New article Practical application of neural networks in trading has been published:

In this article, we will consider the main aspects of integration of neural networks and the trading terminal, with the purpose of creating a fully featured trading robot.

Before starting the development of any trading system, answer the following question: On what principles will this system function? We have two fundamental principles: trading flat and trend continuation. We will not consider derivatives from these two systems, such as intraday trading, use of fundamental data and news, trading at market opening time, etc. I came across descriptions of neural network products, in which authors suggested using them to forecast prices, such as stocks, currencies and so on.

1. Chart shows the operation of a neural network trained for price forecast

Author: Andrey Dibrov

[Dmitry Fedoseev]

Наверное, каждый новичок, который приходит на валютный рынок и пытается реально на нем торговать, не имея какой либо системы  – брал в руки лист бумаги и рисовал на нем таблицу понравившихся индикаторов. 

No, not everyone.

And with the use of neural networks, there is not much point in giving them indicator data at all.

[Maxim Dmitrievsky]

I learnt from the article how to get prices from the terminal in code, as well as stochastic. I also learnt that there is a neural network in matlab.

Thank you very much.

[Andrey Dibrov]

This article is of a rather generalised introductory nature. I didn't want to write about it yet, but I realised that it is necessary. At this stage, I am preparing, let's say, a "mini constructor", which will include fully trained neural network modules for several currency pairs and an Expert Advisor skeleton. All this will be working without any demos and free of charge. If it will be interesting to someone, will help to improve the system both in terms of training neural networks and trading robot - I am ready to share my developments in terms of cooperation.

[Andrey Dibrov]

Dmitry Fedoseev:

No, not everyone.

And with the use of neural networks, there is not much point in giving them indicator data at all.

Of course not everyone. Some people just push buttons.

As far as neural networks go. We do not offend our own personal neural network, which we have in our skull - we do not offend only with the multiplication table.

[Boris Egorov]

The article is about nothing at all

a lot of simple questions without answers .... what is the structure of the neural network and why is it like this, how was the training done and on the basis of what considerations, and what will happen in moments of high volatility when the price jumps down up in fractions of a second and you are throwing data through the disc ..... So to hell with ....

[Rasoul Mojtahedzadeh]

Maxim Dmitrievsky:

I learnt from the article how to get prices from the terminal in code, as well as stochastic. I also learnt that there is a neural network in Matlab.

Thank you very much.

That's right! :D

[Vladimir Perervenko]

Not to repeat, not to check. And hand-to-hand calculations in Excel are for the 9th grade of high school, but not for a programmer. That's a weird article.

[Forester]

Were the tests conducted on training data(backtest) or on new data (forward)?

[Andrey Dibrov]

elibrarius:
Were the tests conducted on training data (backtest) or on new data (forward)?

On new data, but we get more interesting results with the online test https://youtu.be/N2mF9mnVRMs 

[Forester]

Andrey Dibrov:

On new, but more interesting results we get with the online test https://youtu.be/N2mF9mnVRMs

Cool, of course, but without the possibility to repeat it, the value of the article is purely theoretical. Like - I did it, and you should try it.
A robot signal according to the article would be a good incentive to try.
And some people will be able to repeat it on their own systems, if they know what you input (specific indicators with specific parameters, returns, etc.) and what you teach (I assume the price, but for how long in advance? For 5 minutes, for an hour, for 1 point or for several points in 5 minutes increments, etc.) What is the trading algorithm, for example, when the predicted price exceeds some value. Or something else?