Machine learning in trading: theory, models, practice and algo-trading - page 471
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
Fa and Mishanya are not miners)))
Will do for jerking off on parrots. + LightGBM, + CatBoost.
If you want to get a slightly better cut, put python and everything on the GPU...
Is there a significant increase in front of the normal RF? If it doesn't make much sense, I'll use RF from the terminal to continue using it :) Someone here already wrote that the average quality gain is not very significant (error reduction, or whatever). Still, it's better than a bare MLP without any autoencoders + speed
Of course, the main thing is still the method of application, selection of predictors and other black magic
It is written in Java, and it eats a lot of memory. It works no better and no worse than similar R packages. There is one disadvantage and one advantage - constant improvement without backward compatibility.
It is possible to experiment but I would not recommend it for work (IMHO).
Good luck
I just downloaded it and used it a little to get my bearings. I had some problems compiling with vs15. Installing 17 helped.
I did the manual for myself, now on a different computer. Possible solutions on the githab can be seen -
https://github.com/catboost/catboost/issues
Here's an example -
https://github.com/catboost/catboost/issues/72#issuecomment-323100557
Yeah, and it doesn't compile with vs15. I'll try when I have time with 17.
Good luck
My experiments were the end of them, it was too inconvenient to pull the files from the terminal into the predictor and I was too tired to rewrite them on mql :) There are third-party libs in there that are very hard to rewrite.
Some people wrote in one thread that DLRs don't work in MT5 (or don't work yet)
Forum on trading, automated trading systems and trading strategies testing
How to Integrate MQL and R
Vladimir Perervenko, 2017.08.25 13:43
MT4 works fine with DLL. Five, after a number of improvements, does not accept DLL. The last builds of MT5 though I haven't checked.
I think I need to wait until MT5 will settle down and only then refine DLL.
Good luck
If you have a good Expert Advisor and your trading robot, you don't need to copy everything to MQL, there is nothing wrong with "dragging" the files back and forth. You just have to use RAM-Disk and everything will be done directly from memory. Transfer speed is gigabytes per second. Proven. I even started such a topic about half a year ago.
Forum on trading, automated trading systems and testing trading strategies
RAM Disk.
Yuriy Asaulenko, 2016.04.07 21:59
I thought that a virtual disk in the computer memory could help me to speed up MT exchange with external software files. Among the offers I chose AMD Radeon. If anyone is interested, the link is Radeon™ RAMDisk. There is a free version for up to 4GB.
Installed it without any problems. Can be set to overwrite at shutdown, or reset the contents to HDD. At startup it turns on either automatically or by the user.
ZS Turns out that's what we were discussing there with you).
It was written in one thread that DLLs do not work in MT5 (or do not work yet).
I don't need to rewrite everything in MQL, and there is nothing wrong with "pulling" files back and forth. It's enough to use RAM-Disk and everything will be handled directly through the memory. Transfer speed is gigabytes per second. Proven. Even started such a thread about half a year ago.
SZZY It turns out, that's what we were discussing there with you).
No, I have to prepare files for predictor, take them back, wait for him to calculate them, give them back... I can't even talk about full-fledged optimization of parameters, it's an empty task, if I want to connect genetics or something else... everything should be done in one environment, it saves a lot of time... I wish R doesn't do it himself, I'd go for it then ) but I don't like this whole crutch, when you sit down to write a system... you don't even know what you're going to write and you need to concentrate on your thoughts and you have to concentrate on infrastructure and software... without even understanding the point and whether something will work or not... no, I'm lazy, not a globetrotter :)
And I said...
In the end everything will be back to normal, but just as sometimes old age comes without wisdom, some sometimes do not live up to it.
PS:Nice photo in the image, it corresponds to you, something reminiscent of "fellow" Lenonardo Davinci - Salaya
And I said...
In the end everything will go back to normal, but just as sometimes old age comes without wisdom, some people sometimes don't live long enough.
hehe ) No, I have to prepare files for predictor, take them back, wait for it to calculate them, give them back... No full-fledged optimization of parameters is out of the question, it's a waste of time if I want to use genetics or something else... Everything has to be done in one environment, it saves much time... I wish R doesn't do it himself, I'd go for it then ) but I don't like this whole crutch, when you sit down to write a system... you don't even know what you're going to write and you need to concentrate on your thoughts and you have to concentrate on infrastructure and software... without even understanding the point and whether something will work or not... no, I'm lazy, not a globetrotter :)
Imho, the interaction of different software is not a crutch, but a universal trend. And for good reason, almost all software has API and other types of interaction with third-party software.
We don't have API (, but I think, that the future is not inventing bicycles in MQL, but using capabilities of third-party applications in TS.
Imho, you exaggerate difficulties -"we have to prepare files for predictor, take them back, wait for it to calculate them, return... full-fledged optimization of parameters is out of the question . Maybe, maybe).
Imho, the interaction of different software is not a crutch, but a universal trend. And there is a reason that almost everywhere almost any software has API and other types of interaction with third-party software.
We don't have API (, but I think, that the future is not inventing bicycles in MQL, but using third-party applications in TS.
Imho, you exaggerate difficulties -"the predictor must prepare files, fetch them, wait for their calculation, return... full parameter optimization is out of the question . Maybe, maybe.))
I'm afraid, I'm just lazy)) now I've stopped at my own level, something works and satisfies me now... I'm looking towards fuzzy logic together with NS or RF, I have some ideas... it all exists in MT5 :)
It's all in MT5 :)
PS:Nice photo in the image, it corresponds to you, something reminiscent of "buddy" Lenonardo Davinci - Salaya
Oh, better not start, so that then again did not have to apologize again