Machine learning in trading: theory, models, practice and algo-trading - page 2916
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
So the topic is defunct, there has been nothing new for a long time, no thoughts, no articles.
I think the problem is not in people, people here are advanced, but in MQ's policy .They decided that it will work itself out - all they need to do is throw in tools and don't interfere and it will just go on its own...
But it doesn't. If solutions are born, no one will share them on simple enthusiasm, there are no fools.
MQ are clearly trying to repeat the success of indicators, when it was enough to create conditions and people made 3000 indicators. This will not work in the MO topic, the level of development is quite different here and the cost of working solutions is incomparable with indicators.
MQ will have to invest money in this topic and take the development into their own hands. They need to release their solution and integrate it into MT5, there is no other way. It's expensive and risky, but MO will come to trading one way or another.
Microsoft invested in OpenAI in 2018, while Google was puffing up their cheeks and creating tools for the people (TensorFlow). Now they are pissing boiling water and thinking how to save their search business from ChatGPT. For some reason ChatGPT was not born for free on the forums )))
And it will not move.
For example, I study models with different MO algorithms.
And different algorithms have different structure of the final model in the code.
For example, in nonlinear regression, the record of the formula is 900 characters long.
And in GBoosting the code of the final model reaches 15 000 lines! It's hard to imagine how many characters you get there.
So the software problem, namely the Synthetic Tool Formula window, does not allow entering a large number of characters.
There in the Synthetic Tool Formula window, the input length is limited by a very small number of input characters.
It is not clear how to investigate and test MO models expressed through a formula. How will it ferry?
The tool seems to be there, but limits development. From here there is no labour productivity in checking the models.
And you have to dig through hundreds of them. And each model to form a code in a script, very unproductive because of the large binding of the associated environment,
to then just throw out this code, if the model is not successful.
Also suggested developers to introduce for the Formulas synthetic tool , user functions.
That it was possible to turn any type of model into a custom function and use it in the Synthetic Tool Formula.
This solves the issue of both the length of the model and its structure in the form of code.
But the developers replied that they do not plan changes in custom symbols.
How will it work?
If there are such limitations for productive professional research.
He has his nickname blacked out like when he's banned, not replaced with [Deleted] instead
Well, I remember him making a statement like that here in that vein - the moderators rubbed it out. I don't think it's been a month yet.
And it won't run.
For example, I study models with different MO algorithms.
And the structure of the final model in the code is different for different algorithms.
For example, in nonlinear regression, the formula record is 900 characters long.
And in GBoosting the code of the final model reaches 15 000 lines! It's hard to imagine how many characters you get there.
So the software problem, namely the Synthetic Tool Formula window, does not allow entering a large number of characters.
There in the Synthetic Tool Formula window, the input length is limited by a very small number of input characters.
It is not clear how to investigate and test MO models expressed through a formula. How will it ferry?
The tool seems to be there, but limits development. From here there is no labour productivity in checking the models.
And you have to dig through hundreds of them. And each model to make a code in the script, very unproductive because of the large binding of the associated environment,
that then just throw out this code, if the model is not successful.
Also suggested developers to introduce for the Formulas synthetic tool , user functions.
That it was possible to turn any type of model into a custom function and use it in the Formula of the synthetic tool.
This solves the issue of both the length of the model and its structure in the form of code.
But the developers responded that they do not plan changes in custom symbols.
how will it work?
If there are such limitations for productive professional research.
What algorithms, models and programming language are you talking about?
What algorithms, models, and programming language are you talking about?
Algorithms are common: different kinds of regression/classification. Or GBoosting, RandomForest trees.
These algorithms build different structural models.
If we take regression models, the model structure is not so big and looks like a linear formula with mathematical operations.
But it does not fit into the Formula window of the synthetic tool. Because of the limitation of input symbols in the window.
And if we take models on trees, then the model is already obtained on switch constructions. One switch construct is one tree.
And the more trees are used in the model, the more switch constructs it contains.
Thus, the code of the model for this algorithm can grow up to tens of thousands of lines of code.
And in order to be productive in checking hundreds of models of any algorithm, without describing them through MQL5 script, with MQL5 code environment
it was suggested to developers to introduce user functions in the Formula window of the synthetic tool.
Unfortunately, they did not hear it. Or rather, they heard, but referred to the fact that they do not plan to make changes in the Synthetic Tool Formula .
And perhaps, sometime in the future, they may consider the proposal. Which I doubt.
And models need to be checked now, so we have to pull the whole wrapping of the MQL5 code environment through the MQL5 script.
Which significantly slows down the productivity of model checking in MT5.
What do you mean by productivity of checking?
It is much faster to wrap the model code into a custom function and save it in a file or a specialised window,
and enter a synthetic tool formula using the created custom function with the model, and quickly check it,
taking into account auto-synchronisation of different tools if required , than to describe this model with verbose MQL5 pipelining in the script code.
C and MQL5.
Algorithms are common: different types of regression/classification. Or GBoosting, RandomForest trees.
These algorithms build different structural models.
If we take regression models, the model structure is not so large and looks like a linear formula with mathematical operations.
But it does not fit into the Formula window of the synthetic tool. Because of the limitation of input symbols in the window.
And if we take models on trees, then the model is already obtained on switch constructions. One switch construct is one tree.
And the more trees are used in the model, the more switch constructs it contains.
Thus, the code of the model for this algorithm can grow up to tens of thousands of lines of code.
And in order to be productive in checking hundreds of models of any algorithm, without describing them through MQL5 script, with MQL5 code environment
it was suggested to developers to introduce user functions in the Formula window of the synthetic tool.
Unfortunately, they did not hear it. Or rather, they heard, but referred to the fact that they do not plan to make changes in the Synthetic Tool Formula .
And perhaps, sometime in the future, they may consider the proposal. Which I doubt.
And models need to be checked now, so we have to pull the whole wrapping of the MQL5 code environment through the MQL5 script.
Which significantly slows down the productivity of model checking in MT5.
What do you mean by productivity of checking?
It is much faster to wrap the model code into a custom function and save it in a file or a specialised window,
and enter a synthetic tool formula using the created custom function with the model, and quickly check it,
taking into account auto-synchronisation of different tools if required , than to describe this model with verbose MQL5 pipelining in the script code.
C and MQL5.
Isn't this generated by a chat bot? Very similar. No offence intended.
The price moment has three dimensions - price value, time and volume.
Volume is a lump of price?
The price moment has three dimensions - price value, time and volume.
Volume is a lump of price?
Isn't that chat bot generated? It looks like it. No offence intended.
You asked about algorithms so that you could doubt later whether it is not a chat bot?
A detailed answer was given so that there were no other questions.
Apparently the chat bot did a good job that there were no questions left)).
Volume is the number of contracts concluded at a selected price or price range.
Got it. Not a dummy :)
He has his nickname blacked out like when he's banned, not replaced with [Deleted] written on it
The ban was given somewhere between the 9th and 13th. (look for the last comment)) ) Soon enough. Whether he'll be back is the question. Hopefully, without him, the topic forks have become frequent)))))