Machine learning in trading: theory, models, practice and algo-trading - page 1958

 
Maxim Dmitrievsky:

So what did you do there with your seasonalities?

 

I'm not going to lie, but I killed half a day on a complex metric that shows a very important effect. Improving the training and test scores is guaranteed to improve the score on the reference site. Previously, the generalizability of the control area was jumping up and down regardless of the results of training. Now there is a synchronous increase in the generalizing ability both in the training-test section and in the control section.

The formula for the complex metric, if anyone is interested.

General General izability - CCM+ Fscore No thanks!!!!

Лучшая метрика для оценки точности классификационных моделей | DataReview.info
Лучшая метрика для оценки точности классификационных моделей | DataReview.info
  • datareview.info
Оценка качества классификационных моделей — сложная и трудоемкая задача. Сперва аналитик оценивает робастность классификационной модели с помощью таких средств, как AIC-BIC, площадь под ROC-кривой, критерий согласия Колмогорова-Смирнова и др. Следующим логическим шагом является оценка точности модели. Чтобы понять, почему эта задача является...
 
mytarmailS:

So what did you do with your seasonalities?

conditional heteroscedastic seasonalities

 
Maxim Dmitrievsky:

conditional heteroscedastic seasonality

but in worker peasant terms?

 
Maxim Dmitrievsky:

conditional heteroscedastic seasonality

They are both here and there without any rules, also conditional))))

 
mytarmailS:

and in working peasant terms?

depend on the statistics of the previous hours, which are like filters

looks good in pictures, haven't looked at it in practice yet

 
Maxim Dmitrievsky:

depend on the statistics of the previous hours, which are like filters

Looks good in pictures, haven't looked at them in practice yet

From previous hours of the same day, or the same hours of previous days?

 
Maxim Dmitrievsky:

depend on the statistics of the previous hours, which are like filters

It looks good in the pictures, I haven't looked at it in practice yet

I don't understand anything, but I must say that time really has a big effect, I'm experimenting with time here too...

 
mytarmailS:

I did not understand anything, but I will say that the time really affects very much, I also experiment with time ...

repeat it instead of morning prayers

As strong as the AI may be, but the retardation and lack of understanding of 'what to do in general' can only be cured by cramming ))

 
Valeriy Yastremskiy:

From the previous hours of the same day, or the same hours of previous days?

the same

if you look at the signal of one bot from the market that trades seasonals, it has not had a single losing trade in 5 months

If you look at the signal of one market bot, that's a seasonal trade, it has not lost a single deal in 5 months.

Reason: