Machine learning in trading: theory, models, practice and algo-trading - page 2686
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Please add methods for vectors
Engle Granger and Dickey-Fuller tests.
And for matrices Johansen test.
Thanks, we'll take a look and think about the implementation.
Thanks, we'll have a look and think about the implementation.
Thanks. If the solution comes to implementation,
in calculations it is better to use Total Least Squares,
where the result does not change when interchanging time series.
This calculation is more accurate than Least squares.
And since TLS will already be written, it is desirable to add it in a separate method for vectors or matrices, which will return the found coefficients.
Standard TLS for linear model.
There is a calculation for nonlinear model.
Ideally, have both methods in Mql language.
Together, test methods and TLS methods will give a good set of methods for the statistics section.
If we're talking about the audience of consumers and vendors, then of course you need "one secure *.ex5".
If we talk about research for ourselves without a practical possibility to sell the results of our research, then we can use duct tape and make notebooks. We have support for python programs and notebooks right in the terminal and editor.
When it comes to training models, no one is willing to sacrifice speed. Appetite comes with opportunity and resource capacity has long been a limit. That is why there is a hard game of reducing the accuracy of calculations and then correcting the loss of accuracy.
We see the following directions of application of our ML solutions:
Transfer to MQL5 will allow to apply robots in the trading strategies tester and trade without intermediaries.
Correct me if I'm doing something wrong, but it is unrealistic to use MT to work with NS.
First of all, "integration with python" does not help at all. It is easier to run the script separately and transfer data via sockets. Data exchange via "integration with python" is cut down as much as possible. If I have built a vector on the MT side how can I throw it into a python script for polling? And how is it better than sockets?
Quitting TensorFlow and starting to delve into what you're proposing also looks questionable. Yes, it would be cool to compile the solution into .ex5, but that means staying in your sandbox forever. And the atmosphere here is not very friendly:
- I tried to publish articles - I was sent;
- I tried to enter the market with my solution - I was sent (did not pass validation);
- I tried to pull my crypto audience to MetaTrader5 - I faced complete misunderstanding, for them MT5 is a dull retro without normal implementation in the browser and horse brokers' commissions + lack of direct connection to popular exchanges. I had to get away and search for bibles for rendering in the browser myself.
In short, MT5 is a great environment for getting market information and writing bots/indicators (just perfect!), but in all other respects it's a dull bollocks. I don't want to offend anyone, this is my subjective assessment. Now in my project MT5 is an important, but not indispensable piece, little by little I rewrite everything to python and cry that MT5 is much faster and more native, but I have to do it.
If we're talking about the audience of consumers and vendors, then of course you need "one secure *.ex5".
here is an example of crutches that had to be built to stay in MT5, is it normal???
this is a long time outdated version, with outdated models, but I will not update it, there is no point.
If we're talking about the audience of consumers and vendors, then of course you need "one secure *.ex5".
In general, if I understand the market correctly, it is not me, but you who should think about how to turn a solution written in Python + TensorFlow + MT5 into a .ex5 solution that would then be in your marketplace.
... "integration with python" doesn't help at all. It is easier to run the script separately and transfer data via sockets. Data exchange via "integration with python" is cut down as much as possible. If I have built a vector on the MT side, how can I transfer it to the python script for polling?
Switch to R with mt-R. Or maybe there is something similar for python.
Go to R with mt-R. Or maybe there's something similar for python.
Maybe MT5 will come over to my side and I'll just keep doing what I've been doing.
Correct me if I'm doing something wrong, but it is unrealistic to use MT to work with NS at all.
First of all, "integration with python" does not help at all. It is easier to run the script separately and transfer data via sockets. Data exchange via "integration with python" is cut down as much as possible. If I have built a vector on the MT side, how can I throw it into a python script for polling? And how is it better than sockets?
Quitting TensorFlow and starting to delve into what you're proposing also looks questionable. Yes, it would be cool to compile the solution into .ex5, but that means staying in your sandbox forever. And the atmosphere here is not very friendly:
- I tried to publish articles - I was sent;
- I tried to enter the market with my solution - I was sent (did not pass validation);
- I tried to pull my crypto audience to MetaTrader5 - I faced complete misunderstanding, for them MT5 is a dull retro without normal implementation in the browser and horse brokers' commissions + lack of direct connection to popular exchanges. I had to leave and look for bibles for rendering in the browser myself.
In short, MT5 is a great environment for getting market information and writing bots/indicators (just perfect!), but in all other respects it's a dull bollocks. I don't want to offend anyone, this is my subjective assessment. Now in my project MT5 is an important, but not indispensable piece, little by little I rewrite everything to python and cry that MT5 is much faster and more native, but I have to do it.
Support for running *.py files in the terminal is the ability to run already developed models inside the terminal. And the standard MetaTrader 5 for Python library allows you to communicate and trade normally.
It is not for nothing that I pointed out earlier that all third-party integrations are a way of binding with duct tape with a predictable result. We are not interested in such paths at all. So asking to add duct tape is unnecessary, as is skipping a non-piecemeal solution.
As you can see, the MQL5 language is growing very much and we, as its developers, know and publicly declare the goals. Then we achieve them.
Don't worry - almost everything in ML can be done in MQL5. This is exactly the goal we are heading towards.
Over the past 22 years, we have learnt how to achieve big goals.
In general, if I understand correctly how the market works, it's not me, but you who should think about how to turn into .ex5 a solution written in Python + TensorFlow + MT5 that would then be in your marketplace.
Read what I write and what I explain. Read the release notes and published documentation.
Take away your own thoughts and desires. Add the interests of our company, our investments made, our proven track record and our technological capabilities.
Then you will have a clear understanding of what will happen and how.
here is an example of crutches that had to be built to stay in MT5, is it normal???
is a long time ago outdated version, with outdated models, but I will not update, there is no point.
You made a third party solution, protected it as an exe file.
Is it strange that you didn't give away the open source code in Python? After all, you could have easily thrown a *.py script on the chart and everything would have worked?
That's why we put so much effort to make it possible to write complete and protected ML programmes within MQL5.