Machine learning in trading: theory, models, practice and algo-trading - page 2690

 
Vladimir Perervenko #:

Of course it needs to be clarified. For simple models like logistic regression, tree models, etc. it is probably possible. But I am talking about serious models for both TC and tabular data. These two areas have now become very divided and specialised. For tabular data, which are mainly used in machine learning, TabNet(paper, implementations (py) 1, 2, 3) is very promising. And a lot of other packages that give great results. Here is a list of what I have researched and partially use.

Not all of them are used mainly because of machine power limitations and personal preferences. For me training and optimisation for more than an hour is not interesting.

I don't think it will be possible to transfer these models to MCL. And here we can't do without creating infrastructure for linking MKL<->Python.

This is a bit of a digression, but the topic is important to me.

The main idea, I repeat: every developer, whether he is a freelancer, a marketer or a forex/crypto/stock trader, has his "favourite" language and his "favourite" bicycles with crutches to it. We need to share experience of use, not argue what is better. And even more so not to vang about the future of the API.

Good luck to everyone.

Tabular data != time series in the form of tables, they are different things after all

It's not about offence, it's about experience in writing different TCs. Sometimes it's funny to read when a network of 500 layers is attached to one chart of quotes
 
Maxim Dmitrievsky #:
Tabular data != time series as tables, they are different things after all

It's not about offence, it's about experience in writing different TS. Sometimes it is funny to read when a 500 layer network is attached to one quotes chart

Of course, tabular data and time series are different things.

And you cannot argue with it.

Especially hilarious are series of articles like neural network is simple. As an example of programming on MKL - good, but for practice - zero. Well, this is already grumbling.

Good luck to everyone

 
Vladimir Perervenko #:

Of course tabular data and time series are different things.

You can't argue with that.

Especially hilarious is the series of articles like neural network is simple. As an example of programming on MKL - good, but for practice - zero. Well, this is already a grumble.

Good luck to everyone

Read it if you haven't seen it yet.

there's a ranking of classifiers

https://www.timeseriesclassification.com

As far as I remember, neural networks were not in the top of the list
 

TabNet is highly dependent on the dataset and the selected features

Sometimes almost no difference with other classifiers

So I would like more specificity when pushing to apply something. Is this classifier that much superior.

https://arxiv.org/pdf/1908.07442.pdf

I have tried similar architectures for synthesising new data. All neural networks performed worse than GMM on forex time series (less plausible). Whereas on simple tabular data they worked well. I don't remember if there was Tabnet there.

That's why I wrote a clarification that tabular data != time series in the form of tables, the results will be worse.
 
Vladimir Perervenko #:

What are you talking about. Crypto exchanges (specifically Binance) provides...

I thought to poke my nose into the bibles too, but I thought about it and felt sorry for the time....

It's hard for a pythonist when everyone is writing in R in the thread))))
Oh, how so? Python is in the top, how so?)))))) is not the top decides what is better?
 
mytarmailS #:
I thought about poking my nose in the bibles too, but I thought about it and I'm sorry for the time...

It's hard for a pythonist when everyone writes in R in a branch)))
Oh, how come? Python is in the top, how come?)))))) is not the top decides what is better?
There is no SDK, only lefty bibles. Official api only for Java and python. It's all gone again.

What are you writing there? At least show me one solution. Pisuns. Then poke your nose in what you write.

Where is at least one normal integrated solution? One that can be used without tears of pain and emotion 😀

Alexey made a normal solution with catbust, terminal version. I suggested an analogue via python. Everything works in terminal.

There is the same for LGBM
ZY my solutions are used in the market by developers, which kind of hints that they are integrated. And in 2 clicks.


When you ironise, look first at what you have done yourself... and in fact nothing, you have turned packages in R.

You can throw packages at each other on another forum, I guess nothing will change in terms of development of MO in trading.

 

When using R, the product is so cool that it's a shame to sell it 😁

I totally agree that the final product (for mt5) should consist of ex5 file without any integrations and (preferably) without additional files. The history of getting it is not so important - the main thing is that it should work (or be sold).

 
Aleksey Nikolayev #:

Using R makes a product so cool it's just a shame to sell it 😁

I totally agree that the final product (for mt5) should consist of ex5 file without any integrations and (preferably) without additional files. The history of getting it is not so important - the main thing is that it should work (or be sold).

I agree and I think this is the right way. Everything should be in the exe, it should not pull up anything. Otherwise it is not a product for sales.

 
Valeriy Yastremskiy #:

I agree and I think this is the right way. Everything should be in the exe, it should not pull up anything. Otherwise it is not a product for sales.

Tell that to amazon and google. That they do not build their business correctly and their infrastructure is wrong :-)

 
Maxim Kuznetsov #:

Tell that to amazon and google. That they have the wrong business and the wrong infrastructure :-)

It was talking about a specific situation, not everything in the world, and in this specific situation he is right