Machine learning in trading: theory, models, practice and algo-trading - page 2073

[Evgeniy Chumakov]

Aleksey Vyazmikin:

Why would the take point behind the extremum, because the correction could have been 50%, which means that we should take the take point around 100% from the previous interval.

[Aleksey Vyazmikin]

Evgeniy Chumakov:

,

Oh, so you almost drew the system, that I posted earlier as a report :) Only I take TP higher and do not wait till the last interval of ZZ is formed (it is a matter of settings, though).

[Aleksey Vyazmikin]

Alexander Alekseyevich:
So what? Will you throw the code? Or at least a sample.

Sampling link.

Target column "Target_100", at the end of the column with the date the next column and the last two are not used in the training.

The sample is divided into 3 parts, exam.csv is not involved in the training.

[Rorschach]

Evgeniy Chumakov:

As options, an exit before an extremum (with rollover),

Exit at the crossover with the channel based on 3 extrema

[[Deleted]]

There is a separate area - time series classification and related libraries like this

https://pyts.readthedocs.io/en/latest/auto_examples/transformation/plot_rocket.html

Has anyone used it?

[Valeriy Yastremskiy]

Maxim Dmitrievsky:

There is a separate area - time series classification and related libraries like this

https://pyts.readthedocs.io/en/latest/auto_examples/transformation/plot_rocket.html

Has anyone used it?

You should use it, it's an interesting package. Just like a constructor.

[[Deleted]]

Valeriy Yastremskiy:

You should use it, it's an interesting package. It's like a constructor.

I gave link to ROCKET for a reason - it's like cool feature converter. Creates a lot of uncorrelated features from original ones, improves quality of classification.

It is recommended to use it with linear models (because it makes a lot of features).

I will have to test it

[Aleksey Vyazmikin]

Maxim Dmitrievsky:

I gave a link to ROCKET for a reason - it's kind of a cool feature converter. Creates a lot of uncorrelated features from original ones, increases quality of classification.

It is recommended to use it with linear models (because it makes a lot of features).

I will have to test it

Let me know about results - very interesting topic!

[mytarmailS]

Maxim Dmitrievsky:

Creates a lot of uncorrelated features from the original ones,

regular PCA?)

[[Deleted]]

mytarmailS:

regular PCA?)

no