Machine learning in trading: theory, models, practice and algo-trading - page 1165

 
Aleksey Vyazmikin:

And the coefficient for shifting the weights? Like, if the most important 1, then their weight is more, or what?

I don't get it. It's an error on the dataset, how much is classified correctly and how much is incorrect.

 
Maxim Dmitrievsky:

I don't get it. It is an error on the dataset, how many correctly classified how many incorrectly

That's the point, I need to set a priority, for example, I need to strive for 60% of correct zeros and 30% of correct ones, so I want to understand, what formula can set it and express in one figure.

 
Aleksey Vyazmikin:

The thing is, I then need to set a priority, for example, I need to strive for 60% of the correct zeros and 30% of the correct units, that's what I want to understand, what formula can set it and express a single figure.

I don't know, if such a difference it makes sense to have 2 datasets instead of 1

 
Maxim Dmitrievsky:

I don't know, if there is such a difference, it probably makes sense to make two datasets instead of one

And what will give two datasets? In addition, I have little observations - in training only about 9 thousand, and will be 4.5, which is very little.

The description of"BalancedAccuracy" says about some coefficients, I can't understand what they are...

BalancedAccuracy-
use_weights

Use object/group weights to calculate metrics if the specified value is"true" and set all weights to"1 " regardless of the input data if the specified value is"false". The default value is "true".

"

 
Aleksey Vyazmikin:

And what will two datasets do? In addition, I have little observations - in training only about 9 thousand, and will be 4.5, which is very little.

The description of"BalancedAccuracy" says about some coefficients, I can't understand what they are...

BalancedAccuracy-
use_weights

Use object/group weights to calculate metrics if the specified value is"true" and set all weights to"1 " regardless of the input data if the specified value is"false". The default value is "true".

"

I have no idea what you do there )))

see formula

TP - true positive, the rest is similar.

Actually, it's just some tool from Yandex you don't want to understand... they're always mimicking Google and xgboost, mediocrity :)

 
Maxim Dmitrievsky:

I have no idea what you do there ))))

see formula

TP - true positive, the rest is similar.

What does"use_weights" have to do with it?

 
Aleksey Vyazmikin:

What does"use_weights" have to do with it?

I have no idea, it's some kind of crap model, just like Yandex in general

 
Maxim Dmitrievsky:

no idea, it's some kind of crap model, just like the whole Yandex

)))

 
Igor Makanu:

Alas, this is exactly what is needed, but how? The essence is the same - just by sending prices for entry we get one more indicator, which is no better and no worse than the standard indicators

1) You need to create an adaptive indicator adequate to the market...

1" Generate adequate market characteristics

2" to regulate in "on-line" mode indicator parameters according to these characteristics

3" find the best parameters of the indicator on (each bar) according to the current market characteristics (each bar) with the help of technical indicators and OCS testing

This is how I see it

 
Maxim Dmitrievsky:

I have no idea, it's some kind of crap model, just like Yandex in general

And in what models can I adjust what I need - the limits of the values of the correct answers?