Machine learning in trading: theory, models, practice and algo-trading - page 1247

[Aleksei Kuznetsov]

mytarmailS:

Maybe true, maybe not, but interesting

http://brokeruforex.ru/mashinnoe-obuchenie-foreks.html

It's the same as here. There are no informative signs, and everyone is trying to pull something out of the available ones using different models.

[mytarmailS]

elibrarius:
It's the same as here. There are no informative signs, and everyone is trying to pull something out of the available ones using different models.

Have you even read it?

[Yuriy Asaulenko]

mytarmailS:

Maybe true, maybe not, but interesting

http://brokeruforex.ru/mashinnoe-obuchenie-foreks.html

Read. I've written 10 times already - MO is good at solving auxiliary problems in conventional PBXs.

Opening))

[Aleksey Vyazmikin]

mytarmailS:

Maybe true, maybe not, but interesting

http://brokeruforex.ru/mashinnoe-obuchenie-foreks.html

The only thing I found interesting was this idea

"

When to turn your strategy on/off

Knowing when to turn a trade on or off can make the difference between success and loss for your strategy. However, determining when to turn off is not an easy task. We will again use the well-known machine learning algorithm, the hidden Markov states model, to determine the market modes at which our strategy is unprofitable and we should stop trading.

We must first decide what we will use to identify the different modes of strategy efficiency. Let's ask ourselves the question - what factors signal that we should stop trading?

Let's try to use two calculations based on the 10 period simple moving average (SMA) of our profit curve. Let's look at two indicators - the change ratio on the 5 period interval and the distance between the current point of the profit curve and the value of the SMA. The change ratio (ROC) should signal that our curve is in a falling trend, and the distance between the lines will give us a sensitive indicator of strategy performance.

Based on these two input values, we will apply a two-state HMM to decide whether to turn off our strategy:

"

But something I don't understand with the implementation, like putting a mask on the balance and then monitoring the drawdown point, ok, but where is the link to the market predictors and by what rules is the reversal - not clear.

[Aleksei Kuznetsov]

Aleksey Vyazmikin:

I only found this idea interesting

"

When to turn your strategy on/off

Knowing when to turn a trade on or off can make the difference between success and loss for your strategy. However, determining when to turn off is not an easy task. We will again use the well-known machine learning algorithm, the hidden Markov states model, to determine the market modes in which our strategy is unprofitable and we should stop trading.

We must first decide what we will use to identify the different modes of strategy efficiency. Let's ask ourselves the question - what factors signal that we should stop trading?

Let's try to use two calculations based on the 10 period simple moving average (SMA) of our profit curve. Let's look at two indicators - the change ratio on the 5 period interval and the distance between the current point of the profit curve and the value of the SMA. The change ratio (ROC) should signal that our curve is in a falling trend, and the distance between the lines will give us a sensitive indicator of strategy performance.

Based on these two input values, we will apply a two-state HMM to decide whether to turn off our strategy:

"

But I don't understand the implementation, like overlaying a mask on the balance and then monitoring the drawdown point, ok, but where is the link to the market predictors and by what rules is the reversal - is not clear.

Yes. Misha's idea was even cooler: if the trade is unsuccessful - invert the signals of VS).

Probably turn it on after the new training with new data, on which the NS started to fail.
Something seems to me that this will happen too often.

[Ivan Negreshniy]

mytarmailS:

Maybe true, maybe not, but interesting

http://brokeruforex.ru/mashinnoe-obuchenie-foreks.html

The article is more for good health than for mercy, except for the CyberCortex advertisement, I wonder whose it is and what kind of package, does anyone know where to look?

P.S. found here, just trolling the developer - you can tell right away that a concrete guy, not a pathetic Vorontsov's rechirikator and freewar chaff user:)
https://www.mql5.com/ru/forum/36549/page11#comment_1335159

[Ivan Negreshniy]

Vizard_:

cps, the interface is cool, I wish I had an example of how it works...

[Ivan Negreshniy]

Vizard_:

Vanya, I don't want to do it.

And now, on the subject, is there any promising promotion?

Just all the links are old and broken there, but there is some here, but apparently not that ...http://www.cybercortex.org/

[Ivan Negreshniy]

Vizard_:

No idea. Since I've been dealing with this stuff for a long time, a lot of things have piled up.
I got this one a long time ago, tried it and forgot. Here's a screenshot of an example of how it works...

Well yes, thank you, most likely he abandoned the topic or moved on to a broader profile...

[Anatolii Zainchkovskii]

I wonder if anyone has done a study on the presence of abnormal tics? What if there are any? What if we draw BP on these abnormal tics and use them as predictors....