Machine learning in trading: theory, models, practice and algo-trading - page 3128

[Uladzimir Izerski]

Maxim Dmitrievsky #:

It means the deck was shuffled badly

bias - variance tradeoff

Maxim, taking into account your statements and knowledge of markets, let me explain to you what the expression "The tail wags thedog " means.

I know that it will cause your anger and discontent. It doesn't scare me.

Let me explain.

The tail of the dog is the graph you see on the left behind the zero bar, but this very da dog will be on the right.

The dog must be understood and respected like the rest of the forum. Then you will have a chance for success. Read books on psychology.)))))))

[Aleksey Vyazmikin]

Maxim Dmitrievsky #:

It means the deck was shuffled badly

bias - variance tradeoff

That's the point, you can't shuffle the deck if there is data drift. You need to predict it, and possibly generate signs taking it into account, if it has a pronounced vector and not just a fluctuation in the range.

Here I discovered an interesting algorithm "Isolation Forest", which theoretically can capture anomalies/outliers in the sample for training and on new data.

In theory, it can be used to filter the original sample and ignore signals when new data arrives, if they will be very different from those on which the training was carried out.

Do you want to work together to figure this out?

You can read more, for example, here.

[Maxim Dmitrievsky]

Aleksey Vyazmikin #:

The point is that you can't shuffle the deck if there is data drift. It is necessary to predict it, and possibly generate signs taking it into account, well, if it has a pronounced vector, and not just a fluctuation in the range.

Here I discovered an interesting algorithm "Isolation Forest", which theoretically can fix anomalies/outliers in the sample for training and on new data.

In theory, it can be used to filter the original sample and ignore signals when new data arrives, if they are very different from the ones on which the training was performed.

Would you like to work together to figure this out?

You can read more, for example, here.

At the stage of determining the bias given its variability, we need to shuffle. For this purpose, cross-fitting is done (analogue of stability according to Sanych). The variability of this bias may not be linear at all, so this problem cannot be solved through simple inferences. I have learnt to partially solve it, but I always want a better solution.

Also looked in the direction of anomalies, but so far kozul is more interesting.

[Uladzimir Izerski]

Maxim Dmitrievsky #:
Ban these patients already, at last )

They are turning the forum into a dump.

If I turned it into a dump in two posts, then you have created a massive rubbish bin in which you are in charge.

Then let's get on topic.

In what time frame do you think the MoD is capable of quality forecasting?

[Aleksey Vyazmikin]

Maxim Dmitrievsky #:
At the stage of defining bias with regard to its variability, one must shuffle. For this purpose, cross-fitting is done (analogue of stability according to Sanych). The variability of this bias may not be linear at all, so this problem cannot be solved through simple inferences. I have learnt to partially solve it, but I always want to do better.

Without detecting the cause, it is not productive to use different popular methods. Therefore, I would like to measure the variability of data not by models, but by individual predictors with an understanding of the cause of the change.

[Uladzimir Izerski]

Maxim Dmitrievsky #:
Shut your mouth.

I realise you have a problem with predictive capability, but then what are you teaching people about MO?

Let's say in the automotive industry there are roads that MO and hardware can rely on, and in the markets behind 0 bar there is a clear road to all sides of the horizon.

If YOU think narrowing or widening the dog tail section will give you an advantage. Not at all.

[Maxim Dmitrievsky]

Aleksey Vyazmikin #:

Without detecting the cause, it is not productive to use various popular methods. That is why I would like to measure the variability of data not by models, but by individual predictors with understanding of the cause of the change.

We can measure each individual predictor in this way. There is no limit to imagination. It's just matstat and MO, you get what you apply.

Try anomalies, it's easier. I will not explain any more about kozul to those who have not read anything.

[Uladzimir Izerski]

Aleksey Vyazmikin #:

Without detecting the cause, it is not productive to use various popular methods. Therefore, we would like to measure the variability of data not by models, but by individual predictors with understanding of the cause of the change.

It is important to use the detector correctly. This is the basis of the movement.

P.s.

Different factors can serve as a detector, not necessarily of technical nature, but also in combination with FA, news, rumours, etc.

If you are interested, I will give you a hint at the right moment, of course, for free)))).

[Sergey Golubev]

If a user disagrees with some theory being discussed (or the topic/specifics of a thread), and if that disagreement is for more than one/three posts, then I strongly recommend doing the following:

• Create your own thread.
• In the first post of the thread - outline the rules of the thread (what is discussed, and what is not, and how it is discussed, and so on).
• If everything is fine - moderators will monitor the branch according to the rules of the branch.

I understand that making some posts in a very popular and promoted thread is easier than creating your own from scratch and making it popular.
But this is the only way to develop different aspects here without "touching each other".
--------------------

For information.

[Aleksey Vyazmikin]

Maxim Dmitrievsky #:
You can measure each individual like this. There's no limit to your imagination. It's just matstat and MO, so you get what you get.

Try anomalies, it's easier. I won't explain any more about kozul to those who haven't read anything.

I think it's not your place to give me advice.