Machine learning in trading: theory, models, practice and algo-trading - page 1472
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
I spent half a fucking day rewriting my overgrown rl lib for oversampling, but finally finished and got this... the first time (oversampling is automatic), while before I had to choose from a list of models
I got nothing with the meta markup, it's probably just another book nonsense, or my bad hands/thoughts
But of course such nice-looking curves like "we train them in a month and then we get unrestrained profit for several years", as it was shown here recently, has not worked out yet.
what is oversampling? Translated as oversampling. Yandex gives out about audio processing.
There's an article I posted yesterday, before the meta-labeling, and a link to the literature
quote from there
We deal with this by applying the synthetic minority over-sampling technique to create a new training dataset where the label counts are roughly equal
There's an article I posted yesterday, before the meta-labeling, and a link to the literature
quote from there
We deal with this by applying the synthetic minority over-sampling technique to create a new training dataset where the label counts are roughly equal
So, this is just balancing by classes?
It would be clearer in Russian)Ie it's just balancing classes?
In Russian all the same would be clearer)I do not even know how to translate everything into Russian, I already think in these words, because I read everything in English.
I made balancing long ago, but I don't use it.
I'm not sure that full copies (when balancing by duplication) of strings is a good thing. Since this is not new information.
You can still thin the larger class though. But if the difference between classes is big, then 2-5 times thinning is also somehow too much.
Balancing I made myself a long time ago, but something does not use.
I'm not sure that the full copies (when balancing by duplication) strings is good. Since it is not new information.
You can still thin the larger class though. But if the difference between classes is large, then 2-5 times thinning is also kind of overkill.
They also make fictitious examples, and there are all sorts of techniques that you can't figure out without a bottle.
They make fictitious examples there, in short, there are many techniques, you can't figure it out without a bottle
Hello,
I need help to create two trading strategies from those that are now made for training and testing, on python with ZeroMQ (or if it's easier to do otherwise, I'm waiting for suggestions) for trading on metatrader. And also set them on VPS for trading.
On the freelance I applied, it did not work out anything.
I propose to perform this task in the format of online classes, so I understand what and how to do.
I'm sure that a couple of hours, it is definitely possible to do. Remuneration 10 000 rubles. Payment of 2500 rubles I can produce during the class at the end of each hour. And if we finish early, which is very likely, immediately make payment of the remainder.
Send me a personal message if you're interested in my offer! =)
Sincerely,
Sergey
Once again I tried to add real volumes from CME on EURUSD on M1. They only make the forecast worse.
I do not understand why.
It would seem to be additional information, which is directly related to the price movement. But no!
The situation is the same with deltas.
Once again I tried to add real volumes from CME on EURUSD on M1. They only make the forecast worse.
I do not understand why.
It would seem to be additional information, which is directly related to the price movement. But no!
The situation is the same with deltas.
:))) One must work with time - between ticks, and when thinning a series - with time between events. Time in the market is the Grail.