Machine learning in trading: theory, models, practice and algo-trading - page 2943
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
How to do it correctly is another question. I'm still about how feasible - i.e. connect the trained model, but how to do preprocessing then. It is assumed that there is no python after exporting the model.
The model should receive raw input data and do its preprocessing by itself. The concept of pipeline is invented for this purpose. One of the features of ONNX format, for example, is that you can cram the whole pipeline into one file.
The model should receive raw input data and do its preprocessing by itself. For this purpose, the concept of pipeline was invented. One of the features of ONNX format, for example, is that you can cram the whole pipeline into one file.
Shove into the model calculation of zigzags, MAs and other standard and non-standard indicators and duplicate them with different settings? ONNX can do at least one of the ZZ variants, and how many indicators does it have, at least standard ones?
Yep....
Then you don't need MT. You can use API to work - there are only prices and trade commands there too.
The input is only price data. Processing them in one place, not in different places. In the terminal and in training packages it is not quite right. Historically, of course, it is more convenient to take what is processed in the terminal into training packages with prices, but this is a dead-end way. Indicators are of course convenient with their pre-calculation, not recalculation, but it seems to be even set in the task.
In general, the place where the primary data are obtained and the trading environment is managed is not the right place to use for calculations.
Shove into the model calculation of zigzags, MAs and other standard and non-standard indicators and duplicate them with different settings?
Yeah ...
Then you don't need MT. You can use API to work - there are only prices and trade commands there too.
Either that or make a programming environment in which the Python code is interleaved with mql5 code. The second option seems much less realisable.
Either that, or make a programming environment in which python code is interspersed with mql5 code. The second option seems much less feasible.
I described the situation I need to solve. If you are satisfied with just prices and all transformations in python, how does such a concept imply using model transfer at all? Do you propose to duplicate logic for python and terminal?
Obviously, you need a bridge to work with data, and now it is only possible through a file, but then synchronisation is only through files again, which is fraught with problems.
What is the use of running python in the terminal if data exchange is via file or sockets as it was before.
I have data for model polling prepared by the Expert Advisor, not by python. I.e. data should be transferred, and then get back the results for drawing.
In such a scenario there is no sense to use any integration of MT5 and python.
I have the data completely prepared by MT. And gives the model a matrix for calculations (now as a file, but you can do it as an array via DLL). Everything is already working, and for the sake of it, I don't see any of the options you have proposed as motivation to even consider, much less to waste time on duplicating indicators.
Well, I wrote about variants of data exchange only inside the platform. Of course, in practice we all use external to the platform(files, sockets, etc.) ways of data exchange between data analysis tools and TS.
Shove standard and non-standard indicators into the model to calculate zigzags, MAs, etc. and duplicate them with different settings? ONNX can do at least one of the ZZ variants, and how many indicators does it have, at least standard ones?
Yep....
Then you don't need MT. You can use API to work - there are only prices and trade commands there too.
IMHO the model should only deal with training. Generation of matrices for training should be done by other programmes, for example, indicators and/or MT experts.Well, MT is not for big trades, so you can neglect the risks of the impact of calculation time on trade orders. But the niche of MT, even with these risks, is big)))) People like risks)
Please explain how the following formula is obtained in the algorithm of classification on trees with bousting(a link to PDF is possible):
In all materials that I could find in the Internet, the formula is just magically "taken from the ceiling".
Please explain how the following formula is obtained in the algorithm of classification on trees(you can link to PDF):
In all materials that I could find in the Internet, the formula is just magically "taken from the ceiling".
Hard to say :) The mathematical calculations can be seen in this video
Hard to say :) The maths can be seen in this video
Nah, it's blank. I forgot to mention that it's about bousting.