Discussion of article "Practical application of neural networks in trading. It's time to practice"

New comment
 

New article Practical application of neural networks in trading. It's time to practice has been published:

The article provides a description and instructions for the practical use of neural network modules on the Matlab platform. It also covers the main aspects of creation of a trading system using the neural network module. In order to be able to introduce the complex within one article, I had to modify it so as to combine several neural network module functions in one program.

The below figure shows how the Neural Network Module looks like at the initial launch.

НСМ EURUSD_MT5

  1. The Online block is designed to start and stop neural networks during real trading and when testing in the visual mode.
  2. Information fields with the conditions for the signal line crossing the response line of neural networks, when the Online block is activated. 
  3. "Train" is a demonstration block which is designed for training and "retraining" (?) of neural networks.
  4. Fields for outputting the response values of neural networks. Left - neural network responses; right - the signal line. Lower - current bar, upper - previous bar.
  5. The Offline block is designed for outputting neural network responses in a test sample to an array.
  6. Field for entering the averaging value for the neural network response line when using the Online block. (Signal line period). Editable value.
  7. Blocks "Net1,2,3" — three submodules of networks trained in different segments of a time series. Each block includes two neural networks.
  8. NNM operation end time when using the Online block.
  9. Field for entering the NNM operation period in hours, when using the Online block. Editable value.
  10. Counting the time elapsed since NNM start, when using the Online block.

Author: Andrey Dibrov

 

I'm curious if the author of the article honestly bought MatLab (~1000usd) or pirated it ?

 

Neural networks are a very promising area for financial markets.

Thanks to the author for his work!

 

Did the modules themselves have time to malfunction during the journey? For example EURUSD_MT5.exe.

The client part of MT5 is free for the user. So let the bourgeois change the scheme :-))))

MatLab alone, without additional packages, costs so much:

 
It seems that the trained neural network tends to fit the data.
 
Denis Kirichenko:
Matlab alone, without additional packages, costs this much:

.

It's a waste of money)

 
Where can you download all the files from? Or does the author not provide them? Thanks
 
Denis Dementiev:
Where can you download all the files from? Or does the author not provide them? Thank you.

here

EURUSDData.csv(8009.95 KB)

NWI.mq5(5.02 KB)

Matlab_MT5.mq5(9.36 KB)

Snapshot

 
The files can be downloaded below the video))))
 

The article is rather weak - the author does not describe the architecture of the NS, nor the architecture of the linked application, nor the modern way of organising MT5<->Matlab interaction via Matlab Engine API.

It seems that the value of the article is in the presented MT5-advisors, but in fact their value is zero - the presented method of information exchange via files is clumsy, ancient and slow. It may be suitable for trading TF H1 and above, but even on such TFs the practice causes cases of lagging signals, about which the author frankly writes in the section "Work on errors". But the irony is that the main error in the appearance of lags has not been fixed - the file interface remains the same.

MATLAB Engine API for C++ - MATLAB & Simulink
  • www.mathworks.com
Introduction to the basic design of a C++ engine program. Build engine applications on different platforms. Use this sample code to confirm your development environment for the MATLAB Engine API for C++. The MATLAB Engine C++ API comprises functions, classes, and typedefs in namespace. Start and Connect to Start and connect to a MATLAB session...
 
MetaQuotes:

Published article Practical application of neural networks in trading. Let's move on to practice:

Author: Andrey Dibrov

Very good article
12
New comment