Machine learning in trading: theory, models, practice and algo-trading - page 3118

[Aleksey Nikolayev]

Maxim Dmitrievsky direction of the trade and a meta model that predicts the probability of winning (to trade or not to trade):

Let's call the first model the main model, which divides the feature space into buy/sell with a black line. And the second is the meta model, which divides the total feature space into trade/don't trade (red line).

Now let's imagine another variant, when there are two meta models and each of them divides different feature spaces of BUY and SELL classes into trade/non-trade separately (two red lines).

A purely theoretical question "to think about" is whether the second option is better. And if it is better, why. Please comment.

A request, probably even to Alexei Nikolaev, how to determine the effect of such "intervention". After all, you will get 2 probability distributions of two meta models, which can be compared/evaluated/dispersed.

If from a practical point of view, I agree with Forester's opinion.

If purely from a theoretical point of view, one may not contrast the two approaches. To understand this, you can simply think of the straight red lines in the second figure as parts of a single curved line. Essentially, this simply means that the second option is more flexible and complex, giving it more options (in good and bad senses)

[Maxim Dmitrievsky]

Aleksey Nikolayev #:

If from a practical point of view, I agree with Forester's opinion.

If purely from a theoretical point of view, it is possible not to contrast these two approaches. To understand this, you can simply think of the straight red lines in the second figure as parts of a single curved line. Essentially, this simply means that the second option is more flexible and complex, giving it more options (in good and bad senses)

You may get different biases in two different models because of different distributions of trait values for buy and sell, while one model will count something in common. In general I agree, you won't understand until you try it :)

[Evgeni Gavrilovi]

СанСаныч Фоменко #:

You need a quantitative measure of the strength of the relationship between predictor and target. I have written many times on this forum, made references to R packages, even cited the results of my calculations.

I agree, but sometimes a couple of features enhance the quality of the prediction. Here is a simple example. Daytime warming is affected by the amount of cloud cover and humidity.

Every forecaster knows that with high humidity, even with a cloudless sky, the warming will be less significant than with low humidity. So we need to look at the "relationship" of the signs.

[СанСаныч Фоменко]

Evgeni Gavrilovi #:

I agree, but sometimes a couple of signs can enhance the quality of a prediction. Here's a simple example. Daytime warming is influenced by the amount of cloud cover and air humidity.

Every forecaster knows that with high humidity, even with a cloudless sky, the warming will be less significant than with low humidity. So we need to look at the "relationship" of the signs.

In which model of the MoD is it possible to take this into account?

[Uladzimir Izerski]

You don't filter it, you still get a heh-heh error. MO is in essence a fitting of history, which does not have to be repeated exactly.

News, statements of the world's power brokers will leave the MoD in the dark. Since otherwise the rulers should speak in the direction of the MoD and the news should come out according to the instructions of your MoD. The tail wags the dog (c).

But things aren't so sad if you use market models. There may be less room for accuracy, but a higher probability of seeing the direction and duration of the move.

Well, the fact that you read my posts and follow my hints makes me happy).

[Evgeni Gavrilovi]

СанСаныч Фоменко #:

In which MOE model is it possible to take this into account?

There is a catboost.

model.get_feature_importance(type=catboost.EFstrType.Interaction)

https://catboost.ai/en/docs/concepts/python-reference_catboostclassifier_get_feature_importance

[fxsaber]

Forester #:

I think bulls and bears trade differently. The same euro usually falls quickly and then slowly creeps up. Different behaviour.

Is there a script that shows this difference? I have a slightly different view myself(link to a generalised version).

[Renat Akhtyamov]

Uladzimir Izerski #:

You don't filter it, you still get a heh-heh error. The MoD is essentially a fitting of a story that does not have to be exactly the same.

The news, statements of the world's power brokers will leave the MoD in the dark. Since otherwise the rulers should speak in the direction of the MoD and the news should come out according to the instructions of your MoD. The tail wags the dog (c).

But things aren't so sad if you use market models. There may be less room for accuracy, but a higher probability of seeing the direction and duration of the movement.

Well, the fact that you read my posts and follow my hints makes me happy).

Not only the MO won't help, but the probability won't help either.

Margin rules.

[Forester]

fxsaber #:
Is there a script that shows this difference? I take a slightly different view myself(link to a generalised version).

No, I haven't researched this topic specifically. I remembered a quick fall after a slow growth in the times when I tried to trade manually.
Perhaps it was remembered on emotions because it drained my deposit. I do not exclude that everything is even or even vice versa)))

[Forester]

That's the kind of thing I'm talking about. I got it on the first screen I saw. Yesterday between 3:00 and 4:00 p.m.